“Discover the latest advancements in technology, from innovative gadgets to cutting-edge software solutions. Stay updated on tech trends, reviews, and insights that shape our digital future.”
To catch up on some of the news from the past week, check out our ITPro Podcast episode on the conference here.
And with that, we’ve finished the developer keynote! You can refer back to the rest of this blog for all the latest and stay tuned on the ITPro site for more coverage from Google Cloud Next 2025.
Within the Kanban Board, Densmore can ask Code Assist to add code for specific features. If another team member has changed code and broken something – in this case, Densmore uses Seroter as a negative example – Code Assist can flag the changes to make a fix.
When a developer notices a bug, they can tag Code Assist directly in their messaging app, or add a comment within their bug tracker.
Densmore shows us the Gemini Code Assist Kanban Board, which includes something Google Cloud calls a ‘backpack’ – which contains all context for code, security policies, formats, and even previous feedback.
Rounding us out, we’re welcoming Scott Densmore, senior director, Engineering, Code Assist at Google Cloud, to demo a sneak peek at Google Cloud’s software engineering agent.
To share the visualization with colleagues, Nelson can press a ‘create data app’ button to quickly generate a link to the interactive forecast.
The agent uses a new foundation model called TimesFM, which has been built specifically for forecasting, to produce a table with product IDs and dates, as well as a chart with sales over time.
Within the Colab notebook, Nelson can ask the Gemini data science agent to generate a forecast based on his data.
Here to explain is Jeff Nelson, developer advocate at Google Cloud. Nelson starts with Colab, where we’ll be shown a demo of Google Cloud’s new data science agent in action.
We’re moving on to learning about data agents, Google Cloud’s tools for easily analyzing data.
Gemini can see and make sense of information that isn’t apparent to the human eye, says Wong, showing a video of her basketball throw as an example. She adds that a team of developers recently produced an AI commentator for sport and that X Games is interested in using AI for judging.
DiBattista notes that Gemini is capable of analyzing multiple frames at once to evaluate motion, rather than just snapshots. He stresses that he built the tool in just one week, with no need to build a custom model or handle complex data sets.
To demonstrate the amateur pitch, we’re shown a clip of Seroter throwing a baseball outside Google HQ. The system grades him as a ‘C’, with breakdowns or his arm, balance, and stride & drive.
Via Gemini API, DiBattista created a system that can analyze video and produce text analysis of the pitch in the video – both for pros and amateurs.
(Image credit: Future)
The winner of the Cloud X MLB (TM) Hackathon was Jake DiBattista, who’s here now to tell us all about his project – measuring pitches using MLB high-speed video.
What does all this look like in practice? Wong and Seroter say MLB is using Gemini to measure its 25 million data points per game. Google Cloud ran a hackathon to see what innovative use cases people could come up with for Gemini in sports.
“We’re striving to meet developers where you are,” says Cabrera. “Your team can build great apps using Gemini as your IDE of choice, or you can use Vertex AI Model Garden to call your model of choice. No matter what you use, we’re excited to see what you come up with.”
Within Model Garden, developers can test out the model’s response to questions like “what capabilities can you offer for designing renovation subjects?” and see how it responds to evaluate which one best suits their purpose.
Cabrera says while Gemini is her favorite model, Model Garden on Vertex AI offers a range of models from Meta, Mistral, and Anthropic among others.
We’re really cooking now, as Cabrera moves over to Gemini Copilot to produce unit tests by entering a prompt in Spanish – which it quickly does.
Cabrera wants to make an agent to help with budgets, powered by Gemini 2.5. Moving over to Cursor, Cabrera adds input validation to the agent.
For this demo, Cabrera is using the Windsurf IDE, which is intended to support devs with ‘vibe coding’.
Debi Cabrera, senior developer advocate at Google Cloud is now onstage to show us how developers can use Gemini in their IDE of choice, and then bring their model of choice to Google Cloud for their apps.
Google Cloud is at pains to stress that it does not require devs to use Gemini – with Vertex AI Model Garden, there’s a wide range of models to choose from.
Seroter says that Google Cloud is helping developers with its new Agent2Agent, which not only connects agents together but helps developers discover new agents to connect with in the first place.
Within the tool, Gemini suggests a fix to the problem and Sukumaran can immediately deploy it without having to affect anyone’s access to the agent.
To fix this issue, Sukumaran shows us Cloud Assist Investigations, a new tool for diagnosing problems in infrastructure and massively cutting down on debugging time.
Within Agentspace, Sukumaran asks for information related to ordering, expecting a relevant sub-agent to provide the right response. But instead, we’re presented with an error message.
Once she’s deployed this agent system, she’ll be able to share it within Agentspace, where she can interact with the agent.
Sukumaran creates a multi-agent system, right here in the keynote. This means creating a ‘root agent’ with a number of sub-agents, which will work together to automate a task.
Abirami Sukumaran, developer advocate at Google Cloud, is here to show us how to build agents within Vertex AI using ADK with Gemini.
We’re now learning about Vertex AI Agent Engine, which has recently been made generally available and helps enterprises deploy agents with enterprise-grade security. We’ll also hear about Agentspace, Google Cloud’s new solution for building no-code agents, or for developers to share agents they’ve built with the rest of their company.
The moment of truth comes – and the agent produces a detailed PDF proposal that Hinkelmann can access right within the prompt window.
(Image credit: Future)
The next step is to select the AI model Hinkelmann wants for the agent. Because ADK is model agnostic, Hinkelmann says she could use Llama 4 or another model – but in this case will use Gemini 2.5.
Performing RAG requires accessing information from outside the agent, for which model context protocol (MCP) comes in Handy, Hinkelmann says.
Next, Hinkelmann adds an ‘analyze bulding codes’ tool, which allows the agent to use RAG to check a private dataset for local buildings.
Hinkelmann says agents need instructions, tools, and a model. So to start, she uses Gemini in Vertex AI to create a custom instruction: in this case, taking a customer request and creating a PDF proposal.
Here to demo this is Fran Hinkelmann, developer relations engineering manager at Google Cloud.
Wong and Seroter say Vertex AI’s new Agent Development Kit can create an agent that can verify building codes and go deeper into meeting Bailey’s requirements.
Next up, Seroter wants to know what an agent can do.
“An agent is a service that talks to an AI model to perform a goal-based operation using the tools and context it has,” Wong explains.
Wong asks Bailey to go into more detail on the benefits of long context windows.
“This example is some things like photos, images, and a few sketches,” Bailey says. “But with long context, you’re able to send full videos to use for your projects.”
Bailey asks the model to add two globe pendant lights into the image and within seconds, they’ve been added.
In another tab, we’re shown Bailey has used Gemini to generate a prompt for its image generation capabilities and then used this to produce a concept image for the kitchen. It can produce the image, which is photorealistic, in just a few seconds.
Straight away, the model’s ‘thinking’ box shows the model has considered the floor plan (based on a sketched floor plan Bailey provided) and local regulations and building codes.
To start, the pair ask Gemini 2.0 Flash to generate a very detailed plan for remodeling a 1970s style kitchen. Bailey says the model has 65,000 token output window, which is great for generating long plans.
The two are going to make an AI app to help remodel Bailey’s kitchen, taking into account all the details and laws around doing that.
Gemini is key here, of course. Here to show us how is Paige Bailey, AI developer experience engineer at Google DeepMind, and Logan Kilpatrick, senior product manager at Google DeepMind.
Wong says today’s keynote is all about how Google Cloud can help developers build software, from start to scaling, and a sneak peek at the future of development in Google Cloud.
Here to tell us more is Stephanie Wong, head of developer skills & community at Google Cloud and Richard Seroter, chief evangelist at Google Cloud.
Finally, Gemini underpins all these innovations with its large context window, multimodality, and advanced reasoning.
Next, Google Cloud is helping developers be as productive as possible via Gemini Code Assist and Gemini Cloud Assist.
Here to welcome us to the developer keynote is Brad Calder VP & GM at Google Cloud. He says Google Cloud is innovating in three key areas. First up, helping companies build agents, which can collaborate to achieve goals on behalf of users.
To count us down for the final 30 seconds, we’re being shown numbers generated by Veo 2, i ncluding some truly abstract clips such as a giant 1 blasting off to a planet shaped like a 0.
And we’re off! As with yesterday’s keynote, we’re starting with a sizzle reel – this time all about developers, skills, AI, and production.
We’re now sat in the arena and once again listening to the AI-sampled music of The Meeting Tree onstage, accompanied by abstract visuals generated with Google DeepMind’s video generation model Veo 2.
(Image credit: Future)
There are just 30 minutes to go until the developer keynote. Presented under the subtitle ‘You can just build things’, we’re expecting this session to be all about the ease of deploying AI with Google Cloud – expect to hear lots about Agentspace, automation in Workspace powered by Google Workspace Flows, and Google Cloud’s new infrastructure for training custom AI models.
With the press conference done, all eyes are now on the developer keynote – we’ll be seated and ready to bring you images and updates as they come.
Finally, he adds that Google Cloud has European partnerships with firms such as TIM and Thales, to operate in a supervisory role and provide trust and verification in Europe.
He adds that for customers who are worried about long-term survivability, Google Distributed Cloud runs fully detached with no connection to the internet.
Kurian says that technologically, Google Cloud can prevent this from impacting its customers, because the firm doesn’t have access to its customer’s environments and no ability to reach their encryption keys.
Now another question on tariffs from Techzine – specifically on the potential risk that American companies could be ordered to stop delivering services to European customers.
In response, Kurian says Agentspace arose from an observation that organizations struggle with information searches, particularly across different apps. He adds that the service already has 100 connectors live and 300 connectors in development so people can adopt it without ripping out and replacing anything.
We’ve just had a question on how easy it will be for companies to adopt Agentspace when one’s enterprise has already invested heavily in other AI ecosystems such as Microsoft or Oracle, from Diginomica.
A question on tariffs, now – which have been a repeated talking point throughout the event. Kurian is asked whether Google Cloud is prepared for their impact and in response says the “tariff discussion is an extremely dynamic one,” and that Google has been through many cycles like this including the 2008 financial crisis and the pandemic.
Kurian also said Google is working hard to identify opportunities for renewable energy to power data centers and looking to using nuclear as a source of power for its sites.
“We have done many things over the years to improve the infrastructure – for example, we introduced water cooling many years ago for our processors,” he says.
Asked a question on how Google Cloud is meeting the increased energy demand from data centers for generative AI, Kurian says the cost of inference has decreased 20 times.
He adds there’s a competitive advantage to adopting AI and some of the changes in the past few months have changed the European attitude to the technology.
In response, Brady says that Google Cloud is helping EMEA customers with security and flexibility, which are very important in the region, particularly when it comes to not being locked into long-term contracts.
Now a question on pressure facing the EMEA region from our sister publication TechRadar Pro.
The first question is on the challenge of AI adoption in certain countries, to which Kurian says Google Cloud is working hard on its sovereign cloud capabilities. He also highlights the importance of it allowing companies to use its global technology infrastructure in meeting security requirements.
Kurian begins by highlighting how hard Google Cloud is working to expand across the globe and how it now operates in 42 regions.
Before the developer keynote later on, we’re getting to hear from Thomas Kurian, CEO at Google Cloud, Tara Brady, president EMEA at Google Cloud, and Eduardo Lopez, president Latin America at Google Cloud in a press conference.
(Image credit: Future)
It’s coming up on 8:00 in Las Vegas and we’re back to report on day two of Google Cloud Next 2025. With the developer keynote due to kick off this afternoon, there’s sure to be more detail on all the announcements we’ve heard about so far and more hands-on demos of some of Google Cloud’s newest tools.
If you’ve ever wondered what it’s like on the ground at an event such as Google Cloud Next 2025, this photo gives a good impression. You can see it’s incredibly busy here, with attendees in the thousands entering and exiting each keynote. Google Cloud has a huge range of partners and customers, many of whom will be looking to reaffirm or expand their business relationship to make the most of AI, so the event is thick with meetings, roundtables, and live demos in the expo hall.
(Image credit: Future)
“What an amazing time for all of us to experience and work with these technology advances,” Kurian concludes.
“We at Google Cloud are committed to helping each of you in effect by delivering the leading enterprise-ready, AI-optimized platform with the best infrastructure, leading models, tools, and agents. By offering an open multi-cloud platform and building for interoperability so we can speed up time to value from your AI tests, we are honored to be building this new way to cloud with you.”
And with that, the first keynote of the event comes to a close. We’ll keep bringing you all the updates as they happen live from Las Vegas.
Kurian says Google Cloud is working hard on making its innovations easy to adopt in four key ways:
Better cross-cloud networking.
Hands-on work with ISVs to improve Google Cloud integration.
Working with service partners on agent rollouts.
Offering more sovereign cloud compatibility via Google Cloud.
We’re rounding out now and Kurian is back onstage to bring the keynote to a close.
He acknowledges Google’s recent acquisition of Wiz as evidence of how seriously it takes cybersecurity.
In a demo, Payal Chakravarty shows us how Google Unified Security can detect vulnerabilities in code and extensions used within an enterprise’s environment.
The agentic, autonomous features of the new platform can automatically detect when an AI extension has put sensitive data at risk and flag it to a human in the company’s security team. In addition to providing response advice, it can proactively quarantine the suspicious extension.
Continuing at pace, we’re now welcoming Sandra Joyce, VP, Google Threat Intelligence, to hear about the security announcements Google Cloud is making today.
Chief among these announcements is the new Google Unified Security, the new converged security platform for better visibility and faster threat detection.
Read our detailed write-up on Google Unified Security here.
(Image credit: Future)
We’re moving onto Gemini Code Assist, Google Cloud’s AI pair programmer, which Calder says is already being used by a wide range of enterprises.
Google Cloud is today announcing Gemini Code Assist agents, which can help developers to quickly complete tasks such as the generation of software and documentation, as well as AI testing and code migration.
Via the new Gemini Code Assist Kanban board, developers can interact with agents to get insight into why they’re making the decisions they are and see which tasks they’re still yet to complete.
Calder says that Google Cloud is announcing new agents for every role in the data team.
Data engineering agents, embedded within BigQuery pipelines, can perform data preparation and automate metadata generation.
Meanwhile, data science agents can intelligently select models, flag data anomalies, and clean data to reduce the time teams have to spend manually validating all data.
Finally, Looker conversational analytics allows users to explore data using natural inputs. This will be made available via a new conversational analytics API, now in preview, so data teams can embed this easy question and answer layer into their existing applications.
Imagen 3 and Veo 2 models are coming to Adobe Express, we’re told, as the firm pushes forward on AI-generated content.
Moving onto data agents, we’re now welcoming Brad Calder, VP & GM, Google Cloud, onstage.
He tees up a video showing that Mattel is using Google Cloud’s AI to reduce the need for its teams to manually analyze customer sentment.
“We can instantly identify key issues and trends improving growth, efficiency, and innovation,” says Ynon Kreiz, CEO at Mattel.
“For example, we improved the ride mechanism in the Barbie Dreamhouse elevator.”
We’re back to creative agents – it seems creative output is a major focus for Google Cloud at this year’s event. We’re being told about Wizard of Oz at Sphere again – find the details for that at the start of this live blog.
O’Malley is back onstage to discuss purpose-built agents.
For example, Mercedes Benz is using AI for conversational search and route mapping in a new line of its cars.
In a demo by Patrick Marlow, product manager for Applied AI at Google Cloud, we’re shown how the suite can be used to get instant answers and assistance at a garden store.
Marlow is able to hold petunias he has purchased up to a camera and receive real-time, voice output assistance from the agent. For example, he asks if he’s buying the right fertilizer for the plants and the agent is able to recommend an alternative fertilizer and add it to his cart.
In cases where human assistance is required – such as Marlow asking for a 50% discount on his purchase – the agent escalates to a manager in Salesforce.
(Image credit: Future)
O’Malley says Google Cloud’s Customer Engagement Suite is already helping organizations meet customer knowledge demand.
She gives the example of Verizon, which adopted the Customer Engagement Suite. The firm uses the offering to provide its 28,000 customer assistants with up-to-date data and move customers to resolution even quicker.
O’Malley announces new feaures for Customer Engagement Suite, including human-like voices, integration with CRM systems and popular communications platforms, and the ability to comprehend customer emotions.
Customers are using all kinds of agents to unlock new value in their enterprise environment – but what are these different kinds?
Kurian welcomes Lisa O’Malley, leader of Product Management, Cloud AI at Google, to explain more.
O’Malley says we’ll start with customer agents, showing us a video of how Reddit is using Gemini for Reddit Answers, a new conversational layer on the message board website.
Next, we’re told about how Vertex AI Search is helping healthcare and retail organizations to deliver more relevant results to their customers and boost their conversion rates.
“Agentspace is the only hyperscaler platform on the market that can connect third-party data and tools, and offers interoperability with third-party agent models,” says Weiss.
Here to show us more is Gabe Weiss, Developer Advocate Manager, Google Cloud.
Weiss shows us how he can simply identify potential issues with his business’ customers within Agentspace. Based on this, he can ask for an agent to identify client opportunities in the future. He can then iterate on this prompt by asking for an audio summary of its findings, to be delivered to him every morning – creating an in-depth, analytical agent with a few sentences of code.
Finally, he can ask for the agent to write an email within Agentspace, which once approved is automatically sent via Outlook without him having to open the app himself.
It’s time to talk about agents – sound the klaxon. These advanced AI assistants work to automate tasks autonomously, as Kurian explains.
To hear more about the potential of agents, we’re shown a clip of Marc Benioff, CEO at Salesforce.
(Image credit: Future)
“Right now, we’re really at the start of the biggest shift any of us have ever seen in our careers,” Benioff says.
“That’s why we’re so excited about Agentforce and our expanded partnership with Google. I just love Gemini, I use it every single day whether it’s Gemini inside Agentforce, whether it’s all the integrations between Google and Salesforce.”
Starting today, Kurian announces, customers can scale agents across their environment, deploy ready-made agents, and connect agents together.
This will largely be driven by the Agent Development Kit, a new open source framework for widespread systems of agents interacting with one another.
Agent2Agent, a newly-announced protocol. will allow disparate agents to communicate across enterprise ecosystems regardless of which vendor built them and which framework they are built on.
“This protocol is supported by many leading partners who share a vision to allow agents to work across the agent ecosystem,” Kurian says.
Already, more than 50 partners including Box, Deloitte, Salesforce, and UiPath are working with Google Cloud on the protocol.
Within Google Agentspace, enterprises can have Google-made agents, as well as third-party agents and custom-built agents easily communicate with one another.
Vertex AI provides customers with all of Google’s internally-made models as well as open models such as Meta’s Llama 4.
“With Vertex AI, you can be sure your model has access to the right information at the right time,” he says.
“You can connect any data source or any vector database on any cloud, and announcing today you can build agents directly on your existing NetApp storage without requiring any duplication.”
Kurian adds that Google Cloud has the most comprehensive approach to grounding on the market.
Promising Kurian will crowd-surf at tomorrow’s concert, he welcomes the CEO back onstage.
Kurian moves quickly onto Vertex AI, with a look at how it helps customers.
“Tens of thousands of companies are building with companies in Gemini,” he says, giving examples such as Nokia buiding a tool to speed up application code development, Wayfair updating product attributes five times faster, and Seattle Children’s Hospital making thousands of clinical guidelines searchable by pediatricians.
(Image credit: Future)
Once videos have been generated, the user can fine-tune them with new in-painting controls.
In his ive demo, Bardoliwalla paints around an unwanted stage-hand in a close-up clip of a guitar to seamlessly remove him from the final result.
Next, Bardoliwalla uses Lyria to generate music for the trailer. This can be combined in the platform to create quick clips for advertising and more.
Here to show us all how this works in practice is Nenshad Bardoliwalla, Director, Product Management, Vertex AI, Google Cloud.
We’re told his mission is to create a trailer for the party to end the event – complete with a gag about Kurian not wanting to be able to sing Chappel Roan but not getting permission.
Bardoliwalla opens Vertex Media Studio, in which he can ask for a drone shot of the Vegas skyline and choose specific settings such as frame rate video length.
(Image credit: Future)
Onto some more of that creative content we had tee’d up with the DJ (you see, we said it might come up again).
Kurian highlights Imagen 3, the firm’s image generation model, as well as Veo 2, its video generation model. The latter is now capable of adding new elements into filmed video and producing videos that mimic specific lens types and camera movements.
Finally, we’re also told that Lyria is now available on Google Cloud. The model can turn text prompts into short music outputs – the first tool of its kind in the cloud, Kurian says.
Kurian is back onstage, reminiscing on the large progress Google Cloud made last year with Gemini’s multimodality and large, two million token context window.
Gemini is now included in all Google Workspace subscriptions and Kurian tees up a video to show us how businesses are making good use of the service already. In the video, customers say that Gemini is already cutting down their toil and opening new time for valuable work.
Google Cloud’s close relationship with Nvidia runs throughout its hardware announcements today. To hear more, we’re being shown a video of Jensen Huang.
(Image credit: Future)
Huang describes the Google Distributed Cloud as “utterly gigantic”.
“Google Distributed Cloud with Gemini and Nvidia are going to bring state-of-the-art AI to the world’s regulated industries and countries,” he says.
“Now, if you can’t come to the cloud, Google Cloud will bring AI to you.”
Vahdat runs through the core infrastructure announcements from today including Ironwood, AI Hypercomputer, and data storage announcements. As a reminder, you can read about these in detail announcements here.
It’s not all about running workloads in the cloud, Vahdat says. Google Cloud is also announcing Gemini on Google Distributed Cloud, which allows firms to run Gemini locally – including in air-gapped environments.
This opens the door to government organizations using AI in secret and top secret environments.
With that, Pichai is off and Kurian is back onstage.
He explains how Google Cloud is uniquely positioned to support customers, with a massive range of enterprise tools to build AI agents and an open multi-cloud platform for connecting AI to one’s existing databases.
“Google Cloud offers an enterprise-ready, AI platform built for interoperability,” he says.
“It enables you to adopt AI deeply while addressing the evolving oncerns around sovereignty, security, privacy, and regulatory requirements.”
Finally, Google Cloud’s infrastructure is core to its advantages for customers. To help illustrate this point, Kurian welcomes Amin Vahdat, VP, ML, Systems and Cloud AI Google Cloud to the stage.
It’s always good to hear directly from a customer about how AI is helping their business.
We’ve just been shown a reel from McDonald’s, in which Chris Kempczinski, CEO at McDonald’s, explained how AI can be used to predict when machines will need maintenance in McDonald’s restaurants or provide workers with quick answers to their questions.
The announcements are coming fast here in the arena. Pichai rattles off stats about Gemini 2.5, the firm’s new thinking model which is currently the top-ranked chatbot in the world per the Chatbot Arena Leaderboard.
He also notes that Gemini 2.5 Flash, Google Cloud’s low cost, low latency model that allows organizations to balance reasoning with budget for every output.
Pichai draws a direct line between Ironwood and Google’s quantum chip Willow, which it announced last year.
Both are used as examples of the boundaries Google is pushing within its hardware teams, as well as in divisions such as Google DeepMind to crack problems such as weather prediction.
Next, Pichai announces Google Cloud’s 7th generation TPU, Ironwood which brings sizeable performance and efficiency improvements over its predecessors.
A few key stats about Ironwood: it’s capable of 42.5 exaflops of performance, 24 times the per-pod performance of the world’s fastest supercomputer El Capitan.
Read more in our full coverage of Ironwood here.
First off, Pichai says that Google will make $75 billion investment in capital investment in 2025, directed toward servers and data centers.
To further support its AI-hungry customers, Pichai announces that Google Cloud will make its global network available to Cloud customers via Cloud WAN, a new managed solution for connecting enterprises across a wide area network.
“This builds on a legacy of opening up our technical infrastructure for others to use,” Pichai says.
To give the crowd a taste of what AI can do, Kurian welcomes Sundar Pichai, CEO at Google, to the stage.
Pichai opens by paying tribute to The Wizard of Oz at Sphere and then moves on to make some announcements.
(Image credit: Future)
Now the keynote proper begins, with Thomas Kurian, CEO at Google Cloud, taking to the stage to kick us off.
“Google’s AI momentum is exciting – we’re seeing more than four million developers using Gemini, a 20 times increase in Vertex AI,” says Kurian, noting that the firm processes more than 2 billion AI requests per month in workspace, driven by businesses.
Today’s sizzle reel is peppered with AI-generated video, in a show of sophistication by Google Cloud.
(Image credit: Future)
And we’re off! To begin with, as is normal for keynotes, we’re being shown a sizzle reel of Google Cloud’s impact on the industry and hyping up the potential for AI in the enterprise.
Just one minute left until the keynote begins in earnest. Stay tuned as we bring it to you live.
The music we’re hearing will apparently be played throughout the entire conference – musical group The Meeting Tree have scored an entire soundtrack for the event, with the theme of AI.
Paired with Google Cloud’s work on The Wizard of Oz (details lower down in the live blog), it’s clear that Google Cloud is eager to show what it can offer to industries that have been more reluctant to adopt AI to date.
There’s a clear need to acknowledge fears that AI could damage the livelihoods of artists. A constant refrain at yesterday’s event at the Sphere was that ideally, AI should be used to empower creatives rather than replace them. In the event yesterday, Google Cloud suggested that new roles could appear in the creative sector as a result of AI breakthroughs – it will be interesting to see if this is expanded upon at all in the keynote.
We’re now learning a bit more about how that music has been made for the event, via a behind-the-scenes video.
Human musicians were first recorded and then their samples were fed into Music AI Sandbox, which could produce audio outputs that the producers can edit, alter, and use as the basis for new noises.
As you can see, there’s a huge amount of foot traffic this morning as we pile into the Michelob Ultra Arena at Mandalay Bay. As is usual for tech conferences, we’re being serenaded by a live DJ inside the arena itself – more unusual is the visuals for this morning’s music, which have been generated entirely with Google DeepMind’s video model Veo 2.
(Image credit: Future)
As a reminder, the theme for this morning’s keynote is ‘The new way to cloud’, with a focus on interoperability, unification, and more intelligent automation through Gemini AI.
(Image credit: Future)
Last night, we were given a glimpse into what to expect this week at the Sphere, with preview speeches from Google CEO Sundar Pichai and Google Cloud chief executive Thomas Kurian onstage. You can read all about the goings on from the evening further down the live blog.
We’ve already had a range of big announcements ahead of the opening keynote, including the launch of Google’s new ‘Ironwood’ AI accelerator chip and the launch of Google Unified Security, which aims to drive cloud security capabilities for enterprises and demystify cyber complexity in the cloud.
You can read all about these announcements below:
With that, Kurian officially started Google Cloud Next, with confetti cannons heralding the official start of the event.
“If tonight’s event sets the tone for what we plan to bring you for the next three days, I think it’s safe to say it’s going to be an incredible week,” he said.
(Image credit: Future)
Kurian will be back onstage bright and early tomorrow morning at the opening keynote ‘The new way to cloud’. We’ll be bringing you all the updates from that and throughout the conference, both here and across ITPro so stay right here for all the very latest.
In the meantime, why not read my pre-conference analysis of what Google Cloud can do to set itself apart from competitors at this event and the key story it needs to tell.
Next, it was time to hear from Thomas Kurian, CEO at Google Cloud, and James Dolan, CEO at Sphere Entertainment, on the challenges of bringing The Wizard of Oz to the Sphere.
“I’ve been running companies for 40 years and this is one of the first times I ever felt that I wasn’t a customer – I was a partner,” said Dolan, praising the hands-on collaboration of the Google Cloud, Google DeepMind, and Magnopus teams.
(Image credit: Future)
Kurian noted that a total of twenty different models were needed to bring the Wizard of Oz at Sphere to life, with engineers leveraging Google’s extensive TPU architecture and inventing new techniques to expand and recreate the original film frames. This was an enormous technical challenge, not least because the scale and resolution of the screen makes it hard to hide any mistakes in the final image.
“Most importantly, the camera and this amazing theater here at the Sphere is something that doesn’t exist anywhere else in the world,” he said. “So it’s almost like you were told to do AI and your first project was your PhD thesis.”
After Pichai’s speech, we were treated to an extended video showing the behind the scenes of the project. It included detail on how difficult it is to extend existing video footage to fit the Sphere’s unique aspect ratio and resolution, as well as the complexity of generating entirely new footage of characters when they would otherwise have been offscreen.
Engineers had to work iteratively and study the original plans for the film to recreate the characters without making them generic.
(Image credit: Future)
The final project includes special effects such as wind which is blown on the audience and haptic rumbling under the seats – of which we were given a very interactive example.
After entering the Sphere’s cavernous arena, we were treated to an opening speech by Sundar Pichai, CEO at Google. He played tribute to the efforts of all the engineers and creatives who worked on the project, which required intense research and overcoming numerous technological hurdles. Ultimately, it was created using Google DeepMind’s video generation model Veo 2.
(Image credit: Future)
“We have seen significant improvements: super low latency, incredible video quality, multimodal output, so many things we couldn’t have done with AI even 12 months ago,” Pichai said.
“Beyond the technical capability, it took a whole lot of imagination, creativity, and collaboration. Our goal: giving Dorothy, Toto, and all of these iconic characters new life on a 16k screen in super resolution.”
Good evening from Las Vegas, where choice attendees from the event have just been treated to a sneak peek of a brand new attraction opening at the Sphere in August – The Wizard of Oz at Sphere.
Made in partnership with Warner Bros. Discovery, Google Cloud, and Magnopus, the finished product will run as a multi-sensory, 16k recreation of the original 1939 movie for the Sphere’s 160,000-square-foot screen using Google DeepMind’s video generation models.
What was the first smart home product? One could argue it was the electric garage door opener. The first such openers with a radio-based remote controls came to market way back in 1931, predating the first TV remote by 20 years. Comfort and convenience were the motivation behind all three technological advances. In the case of the garage door, people were looking for a way to get out of their cars and into their homes while avoiding the weather. Given that history, it’s a wonder that it took so long to bring IoT technology to the biggest door in the house.
Well, the good news is that it’s here now. And the better news is that most existing garage door openers can be integrated with the rest of your smart home, greatly reducing the cost of that convenience. What’s more, these products are rapidly improving in both simplicity and capability. Buy one and you’ll not only be able to open and shut the door from anywhere—letting in guests, relatives, or delivery people—you’ll also know whether the door is open or closed in real time.
Why you should trust us
TechHive’s writers and editors have been reviewing smart home products for decades, and they draw on their deep and wide experience to evaluate every new product that comes to market. We install the products we review in our own homes to gain real-world experience as we evaluate how well they can be integrated into existing systems as well as how they perform on their own.
The best smart garage door opener controllers
Best smart garage door controller — Chamberlain myQ Smart Garage Hub (model MYQ-G0401)
Pros
Price remains unbeatable
Attractive styling fits in with the typical garage décor
Plenty of third-party compatibility
Cons
Occasional trouble with disconnects
Still no support for a third garage door (you must buy a second controller)
Price When Reviewed:
$40
Best Prices Today:
$27.61
$29.99
It’s still a no-brainer: Everything we said about the Chamberlain model MYQ-G0301 myQ Smart Garage Door Hub is now true of the Chamberlain model MYQ-G0401: It’s the easiest smart garage door controller to set up, the most functional controller on the market, and, it’s the least expensive on the market—by a wide margin. That said, there’s little reason to upgrade from the previous model. The myQ app is simple to configure and use, and the system supports a small but growing number of smart home ecosystems, including HomeKit. While it isn’t compatible with every opener—check online before you buy—it’s definitively the one to get.
Best smart garage door controller, runner-up — Meross Smart Wi-Fi Garage Door Opener (model MSG100)
Pros
Very inexpensive
Solid performance during our testing
Lots of extra features to ensure you don’t leave the door open
Cons
Wired door sensor adds complexity
Virtually no handholding during installation
Price When Reviewed:
$40
Best Prices Today:
$35.99
$50.98
Meross smart home products have left us with mixed emotions. They’re all inexpensive, but value is defined by more than a price tag. The Meross Smart Wi-Fi Garage Door Opener is on the better end of that scale. If Chamberlain’s product doesn’t fit your needs, this one is worth your consideration. (Note: This device is not HomeKit compatible, but Meross offers a separate model that is. It wasn’t available at press time, however, for us to evaluate.)
Read our full
Meross Smart Wi-Fi Garage Door Opener (model MSG100) review
Best security camera for garages — LiftMaster myQ Smart Garage Camera
Pros
Easy to set up and simple to use
Magnetic base makes it tailor-made for garage mounting
Good overall video quality
Integrates well with other myQ gear
Cons
No recording features unless you pay for a subscription
Electrical outlet management can be tricky in a garage setting
Price When Reviewed:
$79.99 (as of June 20, 2024); $149.99 when first reviewed
Best Prices Today:
$106.90
We criticized the high price of this effective special-purpose camera when we first reviewed it several years ago, but times have changed and its MSRP–and especially its street price–have dropped substantially. You don’t need to have a myQ smart garage door controller installed (and the myQ camera doesn’t interact with the myQ controller at all); but if you do, your camera feed will appear directly above your garage door controls within the app. It’s a handy way to get one-stop access to everything that’s going on in the least inviting room of your house. Its magnetic base makes it easy to mount on the bottom of any garage door opener with a metal enclosure. The camera also works with the Key by Amazon system and app, which in this case would empower Amazon delivery drivers to open your garage and place your Amazon packages securely inside.
Read our full
LiftMaster myQ Smart Garage Camera review
How to pick the right smart garage door controller
While garage door openers come in a vast range of brands, styles, and capabilities, the good news is that you’ll likely be able to find a smart controller that works with your system without much trouble.
As I mentioned above, the Chamberlain myQ is my top pick for a variety of reasons, but because it exclusively relies on wireless technology, it isn’t compatible with every system on the market. To start, visit myQ’s compatibility tool and check whether your existing opener is supported. If it is, and you don’t care that it’s not compatible with Alexa or Samsung SmartThings, your work is done: Get the myQ. If it isn’t, you can either get an all-new opener as Chamberlain suggests (although that won’t resolve the Alexa and SmartThings issues), or delve into the world of wired smart garage door controllers.
The Nexx NXG-200 must be attached to your garage door opener via wires, and space can be tight depending on your ceiling height.
Actually, upgrading your old, incompatible door opener is not a terrible idea, and new models are more secure and less expensive than you might think. Considering that a wired garage door controller can run you about $100, it’s worth thinking hard about whether you want to pour more money into an outdated system that might be close to failure, or just upgrade it from the start. (Many new openers have smart technology built in, obviating the need for an add-on controller.)
But if you do have an opener that’s incompatible with our top pick, and you want to keep it around, you’ll need a wired controller like the Nexx Garage NXG-200 (be sure to read our April 5 story about a security vulnerability associated with Nexx garage door controllers) or the Garadget Wired controllers. These must be connected to the opener via a pair of wires, so you’ll need to be comfortable with some minor electrical work in order to install them. Like myQ, Nexx offers an online compatibility tool, but here you’re likely to find that Nexx is either compatible straight out of the box, or compatible only with an additional adapter. In other words, wired controllers are generally compatible with everything, or, at least, I haven’t found any openers yet that aren’t compatible with them.
The Garadget fires a laser once a second at the door to determine whether it is open or closed.As with the product from NEXX, the Garadget must be hardwired to your opener.
The catch involves the adapter. Generally speaking, if you have an older garage door opener, Nexx and Garadget will work with it straight out of the box. If you have a newer opener, you’ll need their adapter as well. This is because newer openers often have a more complex encryption system built in, and a standard push-button remote—which is what wired smart controllers emulate—won’t work with them. The solution is to place a button that is compatible with this encryption in between the controller and the opener: The controller tells the button to activate, which in turn tells the opener to open or close. It’s a little wonky, but in my testing, these setups work just as well as the wireless alternative.
The problem is that it’s just a lot more expensive to do it this way. Purchasing a Nexx and an adapter will run you $105 at press time, and a Garadget plus adapter costs $98. Compare that to the less than $40 you’ll spend on the myQ and there’s really no choice.
Again, if myQ isn’t compatible, either Nexx or Garadget will make for an acceptable alternative, provided you’re willing to spend a little extra to get the job done. We’ll review new products in this space as they come to market and will update our top pick as warranted.
Our GPU benchmarks hierarchy ranks all the current and previous generation graphics cards by performance, and Tom’s Hardware exhaustively benchmarks current and previous generation GPUs, including all of the best graphics cards. Whether it’s playing games, running artificial intelligence workloads like Stable Diffusion, or doing professional video editing, your graphics card typically plays the biggest role in determining performance — even the best CPUs for Gaming take a secondary role.
With 2025, the new GPU releases have been coming fast and furious. Intel kicked things off with the Arc B580 at the end of 2024, followed by Arc B570 in January. Nvidia announced its Blackwell RTX 50-series architecture at CES 2025, then launched the RTX 5090 and RTX 5080 in late January, the RTX 5070 Ti in February. The biggest problem with nearly all of those GPUs? You can’t buy them at MSRP, or sometimes even at all.
The AMD RX 9070 XT at $599 and RX 9070 at $549 arrived on March 6, one day after Nvidia’s RTX 5070 at $549 launched. We’ve tested those cards as well and have updated the GPU hierarchy. Naturally, they sold out as well. It’s not that we don’t want more affordable graphics cards; it’s that AI is all the rage, commanding billions of dollars in sales. It accounted for nearly 90% of Nvidia’s record-setting $130 billion in annual revenue for last fiscal year. Gaming was just under 9% of the total.
A big part of the problem is also that Nvidia and AMD seemingly made some planning errors with the transitions from the prior generation to the new / upcoming hardware. Demand for the prior generation didn’t dry up, and basically all the RTX 40-series GPUs and RX 7000-series GPUs went out of stock, with the exception of the base model RTX 4060 and RX 7600. That started to happen several months ago, leaving a gaping hole of demand for the new GPUs.
Inventory ran out before the new GPUs were ready, creating a backlog of people wanting to upgrade GPUs. Couple that with an inadequate supply of Nvidia’s RTX 50-series cards that have launched and you basically can’t find a reasonably priced graphics card. And then there are new tariffs coming into effect that could further impact prices and availability. It’s a perfect storm hitting the graphics market yet again.
The good news? Some of the latest graphics cards are starting to become available at prices closer to their respective MSRPs. We’ve seen the RTX 5070 stay in stock for hours (yeah, that’s all) at its $549 base price. In fact, as we write this, you can buy the MSI RTX 5070 Shadow 2X OC for $549, direct from MSI. Or you can pay for shipping and get the MSI RTX 5070 Ventus 2X OC from Newegg for $549. Two cards isn’t a habit, but it does represent at least a minor change in availability. The competing AMD RX 9070 by contrast starts at $779 at the time of writing — 42% above the base MSRP, which is more in line with what we’re seeing on a lot of GPUs.
April 2025 Update
We’re in the process of fleshing out the full update with all the latest graphics cards. That includes not only new gaming benchmarks but AI/Professional/Content Creation tests as well. We’ve swapped over the main page to the latest generation GPUs, with some prior generation parts, moving 2024 and earlier results to “legacy” status on the following pages. We’re starting to test older GPUs with the latest test suite as well.
We’re also working on revamping our GPU test suite. We’ve upgraded the test PC to an AMD Ryzen 7 9800X3D and socket AM5 platform, and we’re now in the process of retesting… well, everything, as much as possible. Our latest reviews use the new test suite and test PC, and we now have enough of the latest and previous generation GPUs to swap over to the new data for the hierarchy. There are still a few gaps to fill in (RTX 4060 Ti 16GB/8GB; RX 7900 GRE and 7700 XT), but we’ll get those done soon enough — and then start looking toward testing more of the next older GPUs (RTX 30-series and RX 6000-series).
Our full GPU hierarchy using traditional rendering (aka, rasterization) comes first, and below that we have our ray tracing GPU benchmarks hierarchy. Those of course require a ray tracing capable GPU so only AMD’s RX 7000/6000-series, Intel’s Arc, and Nvidia’s RTX cards are present. The results are all at native resolution, without enabling DLSS, FSR, or XeSS upscaling or frame generation. Below the gaming results, you’ll find our Content Creation Hierarchy, which ranks the GPUs based on their performance in AI benchmarks and professional apps.
Nvidia’s Ada Lovelace architecture powers its previous generation RTX 40-series, with new features like DLSS 3 Frame Generation — with DLSS 4 upgrading the quality of both upscaling and ray reconstruction, plus adding MFG (Multi Frame Generation) on the 50-series GPUs. AMD’s RDNA 3 architecture powers the RX 7000-series, with seven desktop cards filling out the product stack. Intel’s Arc Alchemist architecture brings a third player into the dedicated GPU party, even if it’s more of a competitor to the earlier generation midrange offerings (i.e. RTX 3060 and RX 6650 XT).
On page two, you’ll find our 2024–2022 benchmark suite, which has all the previous generation GPUs running an older test suite and a Core i9-12900K. Page three has an even older 2020–2021 test suite with only rasterization games, running on a Core i9-9900K testbed. The legacy tables are no longer being actively updated. There’s also the “Ancient Legacy GPU hierarchy” (without benchmarks, sorted by theoretical performance) for reference purposes on page three.
The following tables sort everything solely by our performance-based GPU benchmarks, from fastest to slowest. Factors including real-world pricing, graphics card power consumption, overall efficiency, and features aren’t factored into the rankings here. The latest 2025 results use an AMD Ryzen 7 9800X3D testbed. Here are the tables and benchmark results — rasterization games first, then ray tracing games, and finally the content creation results.
Rasterization GPU Benchmarks 2025
For our latest GPU benchmarks, we’re planning on testing nearly every GPU released in the past seven years (the start of the RTX / ray tracing era), and perhaps we’ll extend that even further — depending on how things progress. We test every card at 1080p medium and 1080p ultra; where it makes sense, we also test at 1440p ultra and 4K ultra.
All the scores are scaled relative to the top-ranking 1080p ultra card, which is the GeForce RTX 5090. What else would you expect from a GPU that costs two grand, and often sells at far higher prices?
We have 14 rasterization games we’re using for the latest test suite, though we may swap some of these out in the coming days: Assassin’s Creed Mirage, Baldur’s Gate 3, Black Myth Wukong, Dragon Age: The Veilguard, Final Fantasy XVI, Flight Simulator 2020, Flight Simulator 2024, God of War Ragnarök, Horizon Forbidden West, The Last of Us Part 1, A Plague Tale: Requiem, Spider-Man 2, Stalker 2, Starfield, and Warhammer 40,000: Space Marine 2.
The FPS score is the geometric mean (equal weighting) of all 14 games. Note that the specifications column links directly to our original review for the various GPUs.
Image 1 of 4
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
Rasterization GPU Benchmarks, Key Findings
While the RTX 5090 takes the top spot, pricing is extremely inflated and has been since launch. It’s also largely CPU limited at anything below 4K ultra.
The latest cards typically match previous gen GPUs that are one or two model tiers “higher” (e.g. RTX 5070 roughly matches the RTX 4070 Ti Super, and RX 9070 XT roughly equals the RX 7900 XTX).
The best GPU value in FPS per dollar at 1080p and 1440p is the Arc A750. At 4K ultra, the best FPS/$ comes from the RTX 5070. However, the prices on nearly all of the GPUs are inflated and extremely volatile.
Testing is ongoing and there are quite a few cards we still need to test on the new suite.
Keep in mind that we’re not including any ray tracing or DLSS results in the above table, as we use the same test suite with the same settings on all current and previous generation graphics cards. Only RTX cards support DLSS, with FSR 2/3/4 and XeSS being less supported in general. More importantly, all three upscaling solutions offer differing image quality, and we want to keep things more directly comparable.
One thing that becomes immediately obvious is just how little separation there is between the fastest GPUs at lower resolutions. In some games, there may not even be much of a difference between the 5090 and 4090 at 4K ultra if you’re running in pure rasterization mode. Baldur’s Gate 3 and Assassin’s Creed Mirage for example show a 10% or less delta between the two fastest GPUs. Looking at the overall results, while the 5090 ends up 24% faster than the 4090 at 4K, it’s only 13% faster at 1440p, 5% faster at 1080p ultra, and just 1.5% faster at 1080p medium.
There’s still a sizeable gap between the second place 4090 and the third place 5080 as well. The 4090 leads by 20% at 4K, 13% at 1440p, 12% at 1080p ultra, and 9% at 1080p medium. Looking at AMD’s fastest GPU, the RX 7900 XTX still edges out the 9070 XT, though only by 3–6 percent — with the largest leads again at 4K. AMD’s top solution also comes in slightly behind Nvidia’s RTX 4080 Super, which sits in fourth place overall.
What about Intel? The Arc B580 represents Team Blue’s top solution, but it’s very much a budget to mainstream part. It’s faster than the RTX 4070 and RX 7600 XT, while ostensibly costing less than either of those. However, street prices are currently starting at around $350, which makes it less compelling. We don’t have the 4060 Ti in the charts yet, but based on previous testing we’d expect it to be just slightly faster than the B580 overall — and also quite a bit more expensive. We’ll need to see where Nvidia’s upcoming RTX 5060 and AMD’s RX 9060 land relative to the B580 to declare a budget winner for this generation of GPU architectures.
We’ll hold off on further analysis for now — check the next page to see where older generation GPUs land. Hopefully, prices become a bit more palatable by the time we’ve finished additional tests. But let’s switch over to the ray tracing hierarchy.
Cyberpunk 2077 Phantom Liberty (Image credit: CD Projekt Red)
Ray Tracing GPU Benchmarks 2025
Enabling ray tracing, particularly with the most demanding games, can cause framerates to drop off a cliff. We’re testing with “medium” and “ultra” ray tracing settings. Medium generally means using the medium graphics preset but turning on ray tracing effects (set to “medium” if that’s an option; otherwise, “on”), while ultra turns on all of the RT options at more or less maximum quality.
The four ray tracing games we’re using right now are Avatar: Frontiers of Pandora (a lighter RT title), Cyberpunk 2077 (heavy RT), F1 24 (lighter RT), and Spider-Man: Miles Morales (moderate RT) — all of these use the DirectX 12 / DX12 Ultimate API. The fps score is the geometric mean (equal weighting) of the four games, and the percentage is scaled relative to the fastest GPU in the list, which again is the GeForce RTX 5090.
If you want to see what truly heavy amounts of ray tracing can do, check out our Alan Wake 2 benchmarks where the full path tracing barely manages playable performance even with upscaling on non-Nvidia GPUs. However, and this is a critical point, the number of games where RT truly makes a striking difference in visuals is extremely limited. For most games, we still feel running pure rasterization rendering makes more sense.
Image 1 of 4
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
Ray Tracing GPU Benchmarks, Key Findings and Notes
The RTX 5090 stretches its legs a bit more with RT enabled, as expected.
AMD’s RX 9070 XT outperforms the RX 7900 XTX in our RT tests, indicating AMD’s RDNA 4 enhancements aren’t just fluff claims.
The RX 9070 XT can only manage a tie with the RTX 4070 Ti Super, coming in tenth place overall.
Based on current street prices, the Arc A750 still lays claim to the best bang for the buck at 1080p and 1440p, with the RTX 5070 taking the top spot at 4K.
We are currently running additional tests and had to drop a couple of DXR games from our test suite due to performance issues and game patches. We’ll be adding Control Ultimate Edition back in once testing is complete, along with Indiana Jones and the Great Circle.
Ray Tracing GPU Benchmarks Rankings 2025
Swipe to scroll horizontally
GPU Benchmarks Hierarchy — Ray Tracing Performance
Ray tracing pushes the GPUs a lot harder, depending on the game, which results in a wider performance spread across the GPUs we’ve tested. The 5090 still takes the pole position, with the 4090 in second place, but now there’s at least an 8% gap at 1080p, and that increases to 17% at 1440p and 27% at 4K.
The gap between the 4090 and 5080 is also wider at 1440p and 4K, where the previous generation part is 17% and 24% faster, respectively. That’s without enabling MFG of course, but we wouldn’t take MFG framerates at face value.
AMD’s fastest ray tracing GPU is the new RX 9070 XT, outpacing the 7900 XTX across all tested resolutions. It’s 8–11 percent faster, with the biggest gap at 4K — at resolution where normally the extra memory and bandwidth of the XTX would come into play. But the 9070 XT can only trade blows with the RTX 4070 Ti Super, winning by a tiny amount at 1080p medium and 1440p ultra, and losing by an equally tiny amount at 1080p ultra and 4K ultra.
Intel’s Arc B580 still leads the A770, by a slightly larger margin this time. It’s also generally a better pick than the RTX 4060 due to the lack of VRAM on Nvidia’s card, assuming prices are similar. AMD’s budget RX 7600 and 7600 XT fall off the pace with RT enabled.
Content Creation GPU Benchmarks Rankings 2025
With our new 2025 test suite, we’re also benchmarking every GPU possible in content creation tasks. These consist of AI workloads like Stable Diffusion 1.5/XL, Procyon’s AI Vision suite, MLPerf Client 0.5, SPECworkstation 4.0, and Blender. For gamers, most of these test results are superfluous, but those who plan to use their GPU for other tasks may find the information to be useful.
The individual test results in this suite can swing wildly, depending on the GPU in question, which makes the overall rankings a bit harder to determine. We’ve used the geometric mean of all the benchmarks for the score, but it’s important to check the individual application results for the workloads you plan on running. We’ve included the charts directly below the main table for our Content Creation GPU Hierarchy.
Swipe to scroll horizontally
GPU Benchmarks Hierarchy — AI / Professional / Content Creation Performance
Nvidia’s RTX 50- and 40-series GPUs absolutely dominate in the Stable Diffusion AI workloads as well as in Blender 3D rendering. MLPerf Client shows Nvidia generally in the lead as well, with AMD GPUs doing poorly in the Time to First Token metric, but the Tokens per Second results look much closer than some of the other AI tests.
The Procyon AI Vision suite also heavily favors Nvidia GPUs and even the new Intel Arc Battlemage chips rank ahead of AMD, but AMD’s chips have to run that particular test in FP16 mode, while Intel and Nvidia perform substantially better with optimized INT8 pathways using OpenVINO and TensorRT, respectively.
SPECworkstation 4.0’s GPU inference tests still favor Nvidia, but AMD closes the gap — and Intel falls to the bottom of the list. The Handbrake video transcoding tests only focus on raw throughput, without regards to encoding quality, with Intel’s Battlemage and AMD’s new RDNA 4 GPUs basically tied at the top, then there’s a step down to the AMD RDNA 3 GPUs, and finally all the Nvidia GPUs clump together — though the Blackwell RTX 50-series cards do land slightly ahead of the older 40-series parts.
Finally, we have SPECworkstation’s Viewport test suite, an updated selection of the tests used in SPECviewperf 2020. AMD’s drivers offer more tuning for these tasks, while Nvidia restricts performance for the non-professional GPUs. Certain tests favor Nvidia still (3ds Max and Solidworks) while others (Catia, Creo, Energy, and Medical) heavily favor AMD’s GPUs and drivers. Intel’s drivers are not particularly optimized for these tests and so the Battlemage GPUs are mostly in the bottom two positions — which isn’t entirely unexpected considering they’re also the two least expensive GPUs that we’ve tested in our Content Creation suite.
Test Setup and Hardware for GPU Benchmarks
We’ve used several different PCs for our testing over the years. The latest 2025 update switches to an AMD Zen 5 X3D processor, the Ryzen 7 9800X3D. Here are the specifications for our latest GPU test PC.
Tom’s Hardware 2025 GPU Benchmarks Testbed
AMD Ryzen 7 9800X3D ASRock X670E Taichi Corsair 2x16GB DDR5-6000 CL30 Crucial T700 4TB Corsair HX1500i Platinum Cooler Master MasterLiquid Pro 280 RGB Windows 11 Pro 64-bit 24H2
For each graphics card, we follow the same testing procedure. We run one pass of each benchmark to “warm up” the GPU after launching the game, then run at least two passes at each setting/resolution combination. If the two runs are basically identical (within 0.5% or less difference), we use the faster of the two runs. If there’s more than a small difference, we run the test at least twice more to determine what “normal” performance is supposed to be.
We also look at all the data and check for anomalies, so for example RTX 5070 Ti should be faster than RTX 5070, unless we are CPU bottlenecked, in which case the two GPUs should be basically tied. If we see games where there are clear outliers (i.e. a ‘slower’ card outperforms a ‘faster’ card), we’ll go back and retest whatever cards are showing the anomaly and figure out what the “correct” result would be.
Due to the length of time required for testing each GPU, updated drivers and game patches inevitably will come out that can impact performance. We periodically retest a few sample cards to verify our results are still valid, and if not, we go through and retest the affected game(s) and GPU(s). We may also add games to our test suite over the coming year, if one comes out that is popular and conducive to testing — see our what makes a good game benchmark for our selection criteria.
GPU Benchmarks: Individual Game Charts
The above tables provide a summary of performance, but for those that want to see the individual game charts, for both the standard and ray tracing test suites, we’ve got those as well. We’re only including more recent GPUs in these charts, as otherwise things get very messy.
These charts are up to date as of April 11, 2025.
GPU Benchmarks — 1080p Medium
Image 1 of 21
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
GPU Benchmarks — 1080p Ultra
Image 1 of 21
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
GPU Benchmarks — 1440p Ultra
Image 1 of 21
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
GPU Benchmarks — 4K Ultra
Image 1 of 21
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
GPU Benchmarks — Power, Clocks, and Temperatures
Most of our discussion has focused on performance, but for those interested in power and other aspects of the GPUs, here are the appropriate charts. We’ll run these from highest to lowest settings, as 4K ultra tends to be the most strenuous workload on most of these GPUs.
Image 1 of 4
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
Image 1 of 4
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
Image 1 of 4
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
If you’re looking for the legacy GPU hierarchy, head over to page two! We moved it to a separate page to help improve load times in our CMS as well as for the main website. And if you’re looking to comment on the GPU benchmarks hierarchy, head over to our forums and join the discussion!
Choosing a Graphics Card
Choosing a Graphics Card
Which graphics card do you need? To help you decide, we created this GPU benchmarks hierarchy consisting of dozens of GPUs from the past four generations of hardware. Not surprisingly, the fastest cards are using the latest GPU architectures, though they’re not always a major upgrade over the prior generation.
Of course it’s not just about playing games. Many applications use the GPU for other work, and we cover professional GPU benchmarks in our full GPU reviews. But a good graphics card for gaming will typically do equally well in complex GPU computational workloads. Buy one of the top cards and you can run games at high resolutions and frame rates with the effects turned all the way up, and you’ll be able to do content creation work as needed. Drop down to the middle and lower portions of the list and you’ll need to start dialing down the settings to get acceptable performance in regular game play and GPU benchmarks.
If your main goal is gaming, you can’t forget about the CPU. Getting the best possible gaming GPU won’t help you much if your CPU is underpowered and/or out of date. So be sure to check out the Best CPUs for gaming page, as well as our CPU Benchmarks Hierarchy to make sure you have the right CPU for the level of gaming you’re looking to achieve.
Enthusiasts have been pushing the limits of silicon for as long as microprocessors have existed. Early overclocking endeavors involved soldering and replacing crystal clock oscillators, but that practice quickly evolved into adjusting system bus speeds using motherboard DIP switches and jumpers.
Internal clock multipliers were eventually introduced, but it didn’t take long for those to be locked down, as unscrupulous sellers began removing official frequency ratings and rebranding chips with their own faster markings. System buses and dividers became the primary tuning tools for most users, while ultra-enthusiasts went further – physically altering electrical specifications through hard modding.
Eventually, unlocked multipliers made a comeback, ushering in an era defined by BIOS-level overclocking and increasingly sophisticated software tuning tools. Over the past decade, however, traditional overclocking has become more constrained. Improved factory binning, aggressive turbo boost algorithms, and thermal ceilings mean that modern CPUs often operate near their peak potential right out of the box.
With system bus speeds now tightly controlled to maintain platform stability, most modern overclocking is done through BIOS menus or manufacturer-provided utilities. The challenge has shifted from hardware hacking to managing thermals, navigating firmware quirks, and winning the silicon lottery – but the spirit of overclocking remains alive, driven by those still chasing free performance at the edge of what’s possible.
Here’s a representative sample of landmark processors revered for their overclocking prowess. Read on!
Note: This TechSpot feature was originally published in December 2014. We continue to update and resurface it periodically because CPU overclocking remains almost as cool today as it was back in the 486DX days.
The Pentium MMX arrived amid the height of retailer shadiness and x86 processor vendors responded by locking upper limits for multipliers. As such, many MMXs relied on bus frequency increases for overclocking. Unlocked MMX processors offered more options for overclockers and the unlocked MXX 233 reigned supreme, though its $594 price was prohibitive for many.
At $407, the MMX 166 was a better value, and when paired with a solid 430TX-based motherboard that had a bus speed of 75MHz out of the box, 225 or 266MHz (3 or 3.5 multi) was within reach. To crack 200MHz, MMX 166s with a locked multiplier would need to use the 83MHz jumper setting if available (2.5 * 83 for 207MHz), although stability and heat build-up at this bus speed were far more problematic, as was the sourcing of quality EDO/SDRAM RAM required to run at this frequency.
Intel 486DX2-40
Release date: March 1992 Stock clock speed: 40MHz and 50MHz Overclocked:66MHz(~65%)
The P24 DX2 486 processors introduced the CPU clock multiplier, doubling the system bus speed, while the system bus frequency itself was configurable via motherboard jumpers or DIP switches. Initially including 20, 25 and 33MHz options (later augmented by 40 and 50MHz models), users had a path to overclocking that didn’t require soldering and replacing the clock crystal oscillator.
Alternatively, you could get the performance of a $799 DX2-66 by purchasing the more modestly priced 486DX2-40 for $400 and raising its default bus speed from 20MHz to 33MHz.
Stability and VLB slot issues at bus speeds over 33MHz meant that overclocking headroom decreased as the base clock rose – to the extent that many Intel DX2-66s wouldn’t overclock at all and the few that did were often limited to 80MHz (2 x 40MHz).
Image source: x86-guide
AMD K6-2 300 / 350
Release date: May 1998 (300) / August 1998 (350) Stock clock speed: 300MHz / 350MHz Overclocked:400 – 450MHz(~15 – 30%)
Though the K6-2 struggled against Intel’s Pentium II and III in raw performance, it was extremely popular among enthusiasts looking to build affordable PCs. The addition of 3DNow! instructions gave it a slight edge in multimedia and gaming workloads that could leverage SIMD.
Enthusiasts with Super Socket 7 motherboards often pushed the 300 and 350MHz models to 400MHz and beyond. While not revolutionary in terms of clock headroom, the K6-2’s wide availability, low cost, and tweakability made it a memorable part of the overclocking culture of the late 1990s.
Think legendary. The Celeron 300A was largely responsible for reigniting mainstream processor overclocking in the late 90’s through the ease that it could be accomplished. A 50% overclock to 450MHz was as simple as changing the bus speed from its nominal 66MHz to 100MHz. Although some boards topped out at 83.3MHz limiting the OC to 375MHz, motherboards that supported 103MHz FSB could yield 464MHz.
A better chip with a voltage bump could run at the 112MHz FSB setting to produce 504MHz. Remarkably, the 300A could generally reach 450MHz without any additional voltage requirement over the nominal 2.0 volts. The chip’s performance was also aided by having an on-die L2 cache and with a price of $149 it was particularly accessible to system builders.
Also worth mentioning, Intel’s Celeron 366 (Mendocino) famously hit 550MHz with a good cooler and a 100MHz FSB. Right after the 300A, it was part of the golden age of overclocking on the cheap.
AMD’s Thunderbird pencil mod was an overclocker’s dream. AMD locked the voltage and multipliers of its K7 line in an effort to curtail the fraudulent remarking of processors to higher specifications. Overclockers quickly realized that the circuit bridges built into the silicon package held the key to unlocking performance.
Initially, a combination of connecting bridges in the L3, L4, and L6 blocks gave way to the bridging of L1 connections to unlock the multiplier. Bridging L7 connections to alter core voltage was also an option and the process could be as easy as using a soft lead pencil or conductive silver pen.
With AMD’s EV6 system bus being sensitive to overclocking, multiplier overclocking provided results with the Duron leading the way thanks to its lower base core voltage (1.5v versus 1.7 /1.75v), which enabled higher relative overhead to the maximum 1.85v allowed.
For $112 and a little time, the Duron 600 easily approximated the performance of a processor many times its price.
Intel Core 2 Quad Q6600
Release date: January 2007 (B0 rev) / July 2007 (G0 rev) Stock clock speed: 2.4GHz Overclocked:3.4 – 3.6GHz(~46%)
The Core 2 Quad Q6600 achieved an enviable record of longevity and performance value, becoming the de-facto choice for overclockers who wanted a budget quad-core CPU. Once the processor dropped in price from the initial $851 in January 2007, it quickly fell to $530 in May and further pricing realignment in July coincided with the arrival of the G0 revision. At $266, the 2.4GHz quad-core chip was priced identically to the new dual-core 3GHz E6850, a frequency that the earlier B3 revision Q6600 could comfortably eclipse.
The new G0 stepping offered slightly lower power consumption, which translated into the same improvement in overclocking ability, resulting in many users being able to sustain 3.4 to 3.6GHz fairly readily. The introduction of the affordable Intel P35 platform and further Q6600 price cutting through 2008 to $224 (April) and down to $183 (October) provided the opportunity for solid overclocking in the 50% range (9x multiplier x 400MHz FSB for 3.6GHz) on a moderate budget that remained very competitive long after many of its contemporaries had faded.
The Coppermine Pentium III 500E and 550E’s overclockability lie in conservative binning, a low 100MHz Front Side Bus and the processor’s integrated L2 cache. Budget pricing ($239) and the possibility of using older Slot 1 motherboards via Socket 370 to Slot 1 adapters enabled premium performance for a modest outlay.
The 500E could easily be run at 667MHz by selecting the motherboard’s 133MHz FSB BIOS option or by using tape or lacquer to isolate the Slocket’s A14 pin, while 750MHz (150 FSB) and higher were possible on better boards, producing performance equivalent to the $850 Pentium III 800.
However, there were some caveats to overclocking, including that motherboards needed to support AGP and PCI clock dividers (1:2 and 1:4 respectively) to maintain stability for attached components and fast PC133 RAM.
In early 2004 it came to the attention of the overclocking community that the mobile Barton processors featured an unlocked clock multiplier in addition to being binned for low-voltage operation (1.45v compared with the desktop 1.65v). These factors often produced phenomenal overclocking headroom – something lacking in the desktop models.
When the chip’s overclocking potential was publicized, such was the stampede that its price escalated over 30% from the $75 MSRP in a matter of weeks. With a solid nForce2 motherboard, decent cooling and a willingness to push the voltage to 1.8v and higher, a 30 to 40% overclock was often attainable. While the impressive speed bump couldn’t bridge the performance gap to the new Athlon 64s, the Athlon XP-M 2500+ didn’t cost $200 to $400 either.
Featuring the same silicon as AMD’s San Diego-based Athlon 64 processors, the $125 and $183 Socket 939 Opterons enjoyed a significant pricing advantage over the similarly featured Athlon 64 3700+ at $329 and stacked up even better against the $1,000 FX-57.
Like all upwardly locked multiplier processors, the Opteron’s ability was tied directly to the strength of the motherboard being used. Conservative binning of the Opteron server chips allied with a solid overclocking board such as those sporting the nForce4 chipset with HyperTransport frequencies approaching (and exceeding) 300MT/sec would lead to overclocks seldom seen with enterprise-class processors.
With all the Opteron models having roughly the same overclock ceiling, the lowest priced 144 sold out quickly in many markets.
When Intel announced an upper clock multiplier limit and almost non-existent system bus overclocking for its upcoming Sandy Bridge compatible Cougar Point chipsets, it was widely touted as the end of overclocking on Intel platforms. The truth turned out to be that the 2500K and 2600K were premier overclockers requiring minimal effort in time and cooling for stable overclocks in the 30 to 50% range.
Such was the popularity of the 2600K that submissions from this processor accounted for around 28% of all CPU results to HWBot in 2011 and would exceed those of its successor, the 3770K, in 2012. A low cost of $216 plus solid cooling results when paired with either air or water made Intel’s 2500K the de facto standard by which all other consumer CPUs were judged.
The new Nehalem architecture and X58 platform offered enough promise to coax many users from long-lived Core 2 LGA 775 systems. While the flagship i7 965 EE at $1,000 was cheaper than the Core 2 QX9770 by a third, it still represented little in the way of value compared to the i7 920.
Initial C0 revision Bloomfield CPUs earned a reputation for high voltage requirements past 3.6GHz, the following D0 often had the ability to maintain the nominal 1.26v up to 4GHz and an absolute overclock ceiling approaching 4.5GHz for those tempted to turn the voltage closer to 1.5v.
Such was (and is) the 920’s popularity that it represents over a third of HWBot’s overclocking submissions for 64 LGA 1366 processors.
The arrival of the Northwood core was a welcomed sight after the disappointing Williamette, whose voltage and resulting heat stifled serious overclocking for the mainstream. While the higher-clocked P4s offered little if any value against the Athlon XP, the 1.6A at $125 turned a performance deficit into a win with its low base FSB of 100MHz which could easily be increased to 150 for a 2.4GHz clock speed.
The Celeron’s overclock was higher still thanks to a 20x multiplier, although performance was heavily constrained by the meager 128KB L2 cache. Those seeking higher overclocks would need to push the core voltage past 1.6v either through BIOS settings or the wire mod (connecting CPU pins to raise Vcore limits), the latter being largely responsible for the phenomena of S.N.D.S. (Sudden Northwood Death Syndrome), more commonly known as electromigration.
This factor and the 1.6A cannibalizing Intel’s own higher priced models are seen as the motivation for the company to cease sales of the 1.6A barely six months after its introduction in January 2002.
The FX-8350 was a polarizing chip. Based on AMD’s Piledriver architecture, it was marketed as an 8-core processor, though in reality it had four integer modules with dual-core capabilities. While its single-threaded performance lagged behind Intel’s offerings, the FX-8350 became an overclocking favorite due to its unlocked multiplier and incredible thermal headroom.
Many users pushed it past 5GHz with good aftermarket cooling, and it became a cult classic in the enthusiast scene, especially in budget gaming rigs and benchmark competitions.
Overclocking is most often associated with gaming systems, but dual-processor overclocking has maintained a solid following for over a decade. Long before the QX9775 and Intel’s Skulltrail board became the watchwords for performance excess, many enthusiasts sought the budget Xeon LV 1.6.
The Prestonia core was basically the Pentium 4 Northwood with SMP (symmetric multiprocessing) and HyperThreading added as standard features. With the sub-$200 1.6GHz Xeon drawing a frugal 1.274v, overclockers generally couldn’t take advantage of voltage headroom as most boards were voltage-locked. However, simply raising the FSB would net 2.6GHz.
For the more adventurous, three hard mods could yield a 100% overclock (or more!): the U-Wire mod which involved bridging two (1.5v) or three (1.6v) sets of socket pins, the BSEL mod to isolate or break CPU pins and raise the FSB limit to 200MHz, and the vDIMM mod to raise RAM voltage.
Those willing to push the limits of the technology could be rewarded with a 3.2GHz dual processor performance king for around $700 (CPUs, coolers, board, and RAM).
The initial Thoroughbred-A was little more than a die shrink of the previous Palomino and was somewhat disappointing as a final product. The June 2002 introduction of AMD’s Thoroughbred-B was more tuned for the 130nm process and resulted in higher core frequencies along with being more efficient as the ‘B’ revision demonstrated a remarkable overclock ability with minimal if any voltage increases.
Allied with a strong nForce2 chipset motherboard, the $60 XP 1700+ was fully capable of near 2GHz core speed at its default voltage. With an nF2 board capable of pushing the system bus past 200MHz, it was possible to sustain a 40% overclock with a modest 1.7v, eclipsing the performance of AMD’s $397 Athlon XP 2800+ flagship and putting Intel’s Pentium 4 on notice.
Image source: Sysprofile.de
Intel Pentium D 820 / D 805
Release date: May 2005 (D 820) / December 2005 (D 805) Stock clock speed: 2.8GHz / 2.66GHz Overclocked:3.5 – 4.2GHz(~26%)
The Pentium D 820 was something of an anomaly with two single cores on an MCM package for much cheaper than the cheapest dual-core AMD Athlon 64 X2 at $241 and even undercut the single core Athlon 64 3500+ by $30. The Pentium D 820 offered modest performance in no way challenging the Athlon dual, but some considerable overclocking headroom with judicious voltage and a good air or water-cooling system.
The arrival of Intel’s $129 D 805 further endeared the hot Netburst processor to the budget overclocker. A reduction in nominal system bus speed from 200MHz to 133 was offset by the D 805’s 20x clock multiplier, resulting in no reduction in overclocking fun. For those of modest means, a D 805 paired with a solid 945P board and value-orientated RAM held the promise of performance that was the province of a $500 processor-dictated build.
Intel’s E2000 series effectively signaled the end of both the last surviving NetBurst Pentium D and AMD’s dominance in the budget market. Intel would halve the L2 cache of the E4000 series and further hobble performance with a 200MHz (800 FSB) system bus. What Intel didn’t do was remove the Conroe processor’s ability to overclock.
You could hit a 50% overclock with default voltages and the stock cooler by simply raising the bus speed to 300MHz on either an affordable Intel-based P965/P35 board or one with an Nvidia 650i SLI chipset which allowed greater options with cheaper RAM thanks to its non-reliance on memory dividers.
An aftermarket air cooler, voltage adjustment and some luck in the silicon lottery could see the processors at or near a 100% overclock, delivering performance around the level of the E6700 for a fraction of the cost.
The Core i5 750 was Intel’s first mainstream quad-core processor without Hyper-Threading, based on the Nehalem-derived Lynnfield architecture. It featured an integrated memory controller and a DMI link for PCIe directly from the CPU, improving latency and efficiency.
Though it lacked an unlocked multiplier, its base clock (BCLK) was highly tweakable, allowing enthusiasts to reach nearly 4GHz with good cooling. As the first true successor to Core 2 Quad on the LGA1156 platform, it became a gateway chip into the Nehalem era and remained viable for years with overclocking.
AMD Phenom II X2 550 Black Edition (Callisto) / X4 955 Black Edition (Deneb)
The release of AMD’s revised K10.5 architecture during the early months of 2009 marked a resurgence of the company’s strong value proposition. The emergence of the Black Edition processors also added the welcome addition of the unlocked multiplier to facilitate overclocking.
While the eventual clock speed increases weren’t excessive by historical standards, they did go hand in hand with actual performance gains which comfortably lifted them out of the Core 2 Quad shadow. At $100, the 550 Black Edition represented a superlative value if the two disabled cores could be unlocked (the unlocking of the fourth core would be a major selling point for the X3 720 BE), while the outright performance of the $245 955 BE ensured that only Intel’s more expensive X58 platform exceeded its potential.
When Intel’s Conroe architecture arrived in July 2006, most of the attention was focused on the unlocked multiplier X6800, but it was the cheapest fully enabled (4MB L2 cache) chip that stole the show. For $316, the chip cost a full $200 less than the next step up in performance (the E6700) and already provided results that rivaled AMD’s top Athlon 64s.
With stock cooling and default voltages, you could generally rely on the E6600 to hit 2.7 to 3GHz. If you had an aftermarket cooler, motherboard stability was often the limiting factor as system bus speeds flew past 400MHz and edged towards 450. Such was the overclocking potential that the $999 X6800 and $799 Athlon 64 FX-62 looked positively ludicrous when comparing price and performance with the E6600.
Image source: CPU-World
Intel Core 2 Duo E8400 E0 (Wolfdale-6M)
Release date : January 2008 (C0 rev) / July 2008 (E0 rev) Stock clock speed: 3.0GHz Overclocked:4.0 – 4.5GHz(~41%)
The initial January 2008 C0 revision Wolfdale-based E8400 had immediately ensconced itself as an affordable performance overclocking processor. Five months later, the E0 revision brought a much refined voltage requirement. While some C0-step E8400s were capable performers at the 4GHz level, more often than not, the same frequency could be achieved with stock voltage, settings, and cooler with the new revision.
By the time the E0 arrived, pricing had fallen to $149 for the OEM package with a range of very capable P45 and X48 boards able to maintain bus speeds in the vicinity of 500MHz (2000MHz FSB). The continued stability of these 4+GHz systems many years later is a testament to the quality of both the architecture and the chipsets.
Intel Core i7 3930K (Sandy Bridge-E)
Release date: November 2011 Stock clock speed: 3.2GHz Overclocked:4.5GHz+(~40%)
The Core i7 3930K was Intel’s answer to high-end enthusiasts and prosumers in the early 2010s. Featuring 6 cores and 12 threads on the Sandy Bridge-E platform, the chip ran on the LGA 2011 socket and required the X79 chipset.
Thanks to its unlocked multiplier and improved thermals over earlier hex-core offerings, the 3930K routinely achieved overclocks above 4.5GHz with adequate cooling. Paired with quad-channel memory and strong motherboard VRMs, it became a go-to chip for power users, gamers, and video editors, maintaining relevance well into the Skylake era.
Devil’s Canyon was Intel’s response to criticisms of Haswell’s poor thermals. With improved thermal interface material (TIM) and refined power delivery, the 4790K was the highest-clocked stock CPU Intel had ever released at the time.
Overclockers consistently reached between 4.7GHz and 5.0GHz on air or AIO water cooling setups. Its exceptional single-threaded performance and consistent stability made it a mainstay in gaming builds for years, with many systems built on Z97 platforms continuing to deliver long after newer generations had arrived.
Initially thought to be a re-release of the original Ryzen 5 1600, the 1600 AF turned out to be a hidden gem. Manufactured on the newer 12nm Zen+ process rather than the original 14nm, the chip effectively performed like a Ryzen 5 2600 while retaining the original name and lower price tag.
Thanks to improved thermal efficiency and better memory support, overclockers routinely pushed it to nearly 4.2GHz. The 1600 AF became an unbeatable value proposition in 2020, offering six cores and twelve threads at around $85 and dominating the budget enthusiast scene.
More tech enthusiast content at TechSpot:
Read about the history of the microprocessor, the rise, fall and revival of AMD, or the best CPUs you can buy right now. For lighter reading… check out our top 10 tech pranks and computer tips & tricks you should know about.
Apple’s AirTag is perhaps the most popular Bluetooth tracker on the market. It’s a great tool for using Bluetooth technology to help locate your keys, luggage, and more, and no one is immune from misplacing these necessities. While AirTags can be easily attached to keys, cutting-edge AirTag accessories let you use them on pet collars, remotes, and even wallets.
We’ve compared the best AirTag wallets from brands like Ridge, Titan X, and more for card capacity, design, and overall value. We’ve also provided alternative Bluetooth tracker options, like Bluetooth wallet cards. Right now, the best AirTag wallet for most people is the Titan X Trifold AirTag wallet.
What’s the best AirTag wallet right now?
The accessory market is flooded with different AirTag wallets and cash holders. While you may be looking for a Bluetooth tracking card, we tested many options available to ensure you never lose your wallet again. Our pick for the best AirTag wallet overall is the Titan X Trifold AirTag wallet because it offers plenty of room for cards and cash in a functional and sleek design. It’s durable and offers great value. Keep reading to see how our other top AirTag wallets and accessories stack up based on our hands-on testing and recommendations from experts.
Also: Attention Apple AirTag users: Don’t ignore this warning
The best AirTag wallets of 2025
Show less
The Titan X Trifold wallet comes with a built-in AirTag holder, internal cash and card storage, extra pockets, and a handy trigger mechanism to help dispense your most used cards.
Unlike the Ridge wallet, you don’t need to purchase the AirTag holder separately, and the Titan X is slightly cheaper than the Ridge, especially while it is on sale for $65. It’s a durable daily driver and offers great value.
Also:Forget Ridge: I tested an AirTag wallet that’s more functional (and cheaper)
Testing the Titan X wallet for several weeks left me quite impressed. It’s even more valuable than purchasing an option like the Ridge wallet, especially if you already have an AirTag on hand to pop in it. The Titan X wallet is constructed of leather and aluminum and features a traditional trifold design. There is internal cash storage and traditional internal “pockets,” which I found helpful and more secure for my ID and money. The trigger mechanism is a manicure saver.
Plus, it feels uber-premium and comes in a great box perfect for gifting. My only complaint is that the Titan X doesn’t come in more colors, and the internal cash pockets remind me of cheap school folders.
Current customers wrote on the Titan X website that they liked the wallet’s capacity, durability, design, and functionality. The wallet has a 4.5/5 star rating online, and several customers said they were impressed with its quality and construction.
Titan X AirTag Wallet specs: Type: Expandable | Material: Aluminum and leather | RFID blocking | Built-in AirTag holder | Internal pockets and cash storage | Trigger mechanism design | Dimensions: 3.9 x 2.8 x 0.7 inches
Read More
Show Expert TakeShow less
Show less
The Ridge wallet is a slim RFID-blocking wallet designed to hold up to 12 cards. Cash can be stored via a cash strap or money clip on the outside of the wallet, and the brand offers three AirTag configurations for tracking: an AirTag holder designed to or for $25, and the AirTag carbon add-on holder that attaches outside any Ridge wallet for $40.
The Ridge offers durability and functionality while also providing top-notch security and smart capabilities. Although it is a bit pricey, it has a lifetime warranty to keep up with your daily use (and a risk-free 99-day trial period). All things considered, a quality, traditional leather-style wallet could easily run you over the $100 mark, too, which makes this long-term investment worth it, in my opinion.
I gifted this wallet to my Dad a year before I started using one myself, and I can confidently say it’s a winner. Before gifting him the Ridge, he used several other wallets, including a traditional leather and titanium case style, but none fit his needs.
My Dad is a shove-everything-into-your-pockets-and-go kind of person, and no wallet could ever keep all his cards, cash, and receipts organized until the Ridge. The minimalist design helps ensure he can put what he really needs into his wallet and makes it easy to access cards and cash whenever he wants.
I initially thought the necessary maneuvering to slide the cards in and out would be annoying. Still, after a little practice, he said it became second nature and is surprisingly easier than opening and removing cards or cash from a traditional wallet. I agree that a brief learning curve is involved, but once you get it down, it’s a non-issue.
The wallet’s practicality, coupled with the RFID technology, numerous color and metal options, and the addable AirTag tracker, make the Ridge nearly faultless. Although it can get a little chunky when it’s full to the brim with cards (but alas, what wallet doesn’t?), this is a great wallet for small handbags, concert or stadium bags, and going out on the town.
Current Ridge customers wrote on the website that they liked the capacity, durability, design, and functionality of the wallet. The wallet features a 4.8/5 star rating online, and 97% of customers said they’d recommend it to a friend. Several customers said they were repeat purchasers and their wallets stood the test of time.
Ridge Wallet specs: Type: Expandable | Material: Aluminum | RFID blocking | Accessories and add ons available | Several materials and designs | Cash strap or money clip for cash storage | Dimensions: 86 x 54 x 6 mm
Read More
Show Expert TakeShow less
Show less
If you’re not dead set on using an Apple wallet or AirTag accessory to track your wallet, it may be worth considering a MagSafe wallet equipped with location tracking. The Journey LOC8 is the finder version of our pick for best overall MagSafe wallet, the Ezmo wallet (pictured above).
Journey’s MagSafe wallet is the perfect balance of style and function, with enough space to hold five cards. The wallet also doubles as a phone stand, so you can watch your favorite entertainment on the go or keep your phone screen visible at your desk. The stand has adjustable viewing angles in both portrait and landscape modes. It features a protective lining to safeguard your cards’ information with RFID-blocking technology while preventing them from becoming demagnetized.
The LOC8 takes all the great features of the Ezmo wallet one step further, offering location tracking — no AirTag needed. It uses Apple’s Find My technology for easy integration, is rechargeable, and comes with built-in speakers so you can easily ping the LOC8 wallet to find its location.
I tested the Ezmo version of the Journey wallet for several weeks during my commute from New Jersey to New York City, and I like it enough to recommend the LOC8 with built-in tracking confidently. I like the premium feel and how easy the wallet makes it to access my transit cards during the commute. I also like how the wallet provides enough space for five cards and folded cash. It’s perfect for holding my debit card, work badge, and other incidentals like paper train tickets. The phone stand was nice, too.
Although it’s one of the pricier options for a MagSafe wallet, I think it’s well worth the investment due to its multi-functionality and excellent quality, especially if you use your MagSafe wallet every day.
Current users on the Journey website said they liked the slim profile, smart design, larger card capacity, and sleek material — so much so that the wallet has a 5/5 star rating online, with one customer noting that it “does it all really well.”
Journey LOC8 tracking wallet and phone stand features:Type: MagSafe bifold | Multi-functional | RFID protection | Built-in lock to prevent card slippage | Water resistant | Built-in speaker | Apple Find My System compatible | Rechargeable | Dimensions: 66 mm x 103 mm x 14 mm | Card capacity: Up to 5 cards
Read More
Show Expert TakeShow less
Show less
If you like affordable luxury tech products, don’t look any further than Bosca’s minimalist AirTag wallet for all your card storage needs. Part of my job at ZDNET includes testing tons of tech accessories, from MagSafe gear to AirTag accessories and beyond, but this is the first AirTag wallet I’ve tested in a while that I was surprisingly impressed with.
There’s nothing unique about the design of this Bosca wallet. In fact, if you went on Amazon, you’d see tons of similar options at varying price points offering similar features: slim design, a few card slots, maybe an ID window, and, of course, that trusty little spot to tuck your AirTag. The hybrid card slot/wallet combo has become a popular option as many people shift to cardless payment options like Apple Pay and Google Pay more and more and carry fewer cards daily.
But what makes Bosca’s stand-out is the craftsmanship and leather used to construct it. Upon opening this wallet, I knew right away that it was what the kids call legit. The leather is smooth, and the Saddle color has a striking and elevated tone. It looks like something that didn’t fall out of Amazon’s massive factory market, thanks to the full grain, Italian, hand-stained leather and intentional construction.
It holds cards well, has a dedicated ID slot with a slit to make grabbing your identification easy, and has two pockets to carry extra cash, cards, or other incidentals. It’s super easy to snap your AirTag in and be on your way, and most importantly, it’s the kind of tech accessory that feels like it will withstand the test of time — perhaps even getting better with age. I compared the leather of this wallet to some of my personal handbags and small goods from various other brands, and there’s a similar quality standard.
It is pricier than other options on this list (though not as expensive as a Ridge wallet) and I do wish there were perhaps more color options, but for what is offered I feel that this wallet is a confident balance of affordable luxury that justifies its price tag, and could prevent the type of incessant rebuying that can often occur with tech accessories.
Bosca AirTag Wallet features: Type: Hybrid cardholder design | Dimensions: 4.375 H x 3.00 W x 0.125 D | Card capacity: 3 Card Pockets, 2 card/cash pockets, 1 ID slot | Cash storage | Full grain, Italian, Hand-stained leather
Read More
Show Expert TakeShow less
Show less
The Spigen AirTag Wallet S is an affordable and expandable design. Similar in overall construction and function the the Ridge wallet, the main difference is the material and price point.
Despite its thin appearance, the Spigen AirTag Wallet S holds up to 12 cards. The polycarbonate shell makes for a super-light build, so it won’t weigh your bag or pockets down. The nylon elastic cash strap easily stows your bills, and the wallet also incorporates special RFID-blocking protection to prevent card readers from stealing your information.
“The Spigen AirTag Wallet S’s sleek design and compact form initially drew my attention,” says Liz Ho, an experienced camping guide, travel expert, and founder of Camping Guidance. “Its built-in AirTag compartment gave me peace of mind, knowing I could easily locate my wallet if it ever went missing.”
Amazon customers wrote that they liked that the Spigen S wallet was more affordable, had a dedicated, built-in spot for an AirTag, and held both cash and cards. However, some experienced damage early on and didn’t like the plastic material. Overall, customers said they were happy with the wallet’s appearance and value but wished it had some improvements in material and performance.
Spigen AirTag Wallet S specs: Type: Expandable | RFID-blocking | Cash strap for cash storage | Material: Polycarbonate | Cash strap for cashe storage | Dimensions: 1.18 x 3.15 x 1.97 inches
Read More
Show Expert TakeShow less
Show less
The Doeboe AirTag Wallet is the best budget AirTag wallet with a surprisingly lightweight build despite its leather material.
Ricky Allen, marketing director of Ever Wallpaper, says, “It’s a great wallet with all the necessities, like card slots, a pocket for bills, and an AirTag keychain holder.”
Made of Crazy Horse leather, this bifold wallet features 10 card slots and a bill pocket to stash your cash. It also has an ID window, so you don’t have to take your driver’s license out whenever you need to show it. The minimalist design is simple to manage and lightweight, so it won’t weigh your pockets down.
Allen commented on the security as well. “I liked how easy it was to insert my AirTag and the anti-theft RFID protection it provides. Security is very important to me, and I’m glad this wallet provides the added protection of RFID-blocking.”
Current Doeboe wallet users on Amazon said they liked the wallet’s appearance, durability, and compactness, while several customers emphasized its genuine leather construction.
Doeboe AirTag Bifold Wallet specs: Type: Bifold | Material: Leather | ID window | RFID blocking | Dimensions: 1.97 x 1.57 x 0.59 inches
Read More
Show Expert TakeShow less
Also:How to tell if an AirTag is tracking you
What are the tariffs in the US?
President Donald Trump’s new economic plan could shake up global tech manufacturing — potentially hitting consumers the hardest. Imports from some countries may decrease in availability and increase in cost more than others. It may be time to take note of products you were looking to buy in the next few months and accelerate your purchasing timeline.
Small tech accessories like chargers, cables, adapters, MagSafe gear, and more will not necessarily be immune to the current economic times. Many accessories utilize imported metals and materials. However, accessories like AirTag wallets will likely still remain “affordable” in comparison to larger tech products like laptops and iPhones.
We do not encourage panic buying, but if you’re seriously in the market for a tech product, sooner rather than later may be a good idea to consider purchasing new tech before the recoil from tariffs hit.
Also: If you’re planning to upgrade your phone, you might want to buy a new one now – here’s why
The Titan X AirTag Wallet is ZDNET’s top pick for best overall AirTag wallet.
With a slim and stylish, yet easily expandable and durable design, the Titan X wallet handles up to 10 cards, and offers space for cash. Equipped with RFID-blocking technology, the Titan X is the crème de la crème of trackable, practical wallets, and is a true buy-one-time purchase.
See how our other top picks compared to the Titan X wallet in type, cost, and card capacity below:
Best AirTag wallet
Cost
Type
Card capacity
Titan X AirTag Wallet
$80
Expandable
up to 10
Ridge Wallet
Starts at $95
Expandable
up to 12
Journey LOC8 Tracking MagSafe Wallet and Stand
$90
MagSafe bifold
up to 5
Bosca AirTag Wallet
$85
Hybird card holder
up to 6
Spigen AirTag Card Holder Wallet S
$50
Expandable
up to 12
Doeboe AirTag Bilfold Wallet
$26
Bifold
up to 10
*MSRP at the time of writing. Please note that actual prices may vary depending on available sales, deals, discounts, and coupons.
Show more
It isn’t easy to choose the best AirTag wallet when there are so many options on the market. These expert recommendations may help you in the buying process so you can be sure to find the right wallet for your needs.
Choose this AirTag wallet…
If you want…
Titan X AirTag Wallet
The best overall AirTag wallet for function, style, security, and longevity. Easily customize this wallet to fit your lifestyle.
Ridge Wallet
The best splurge AirTag wallet. Easily customize this wallet to fit your lifestyle.
Journey LOC8 Tracking MagSafe Wallet and Stand
A MagSafe wallet option equipped with Apple’s Find My tracking technology.
Bosca AirTag Wallet
An affordable luxury, hybird AirTag wallet.
Spigen AirTag Card Holder Wallet S
The best overall affordable AirTag wallet alternative with great value.
Doeboe AirTag Bifold Wallet
A budget bifold wallet. This best AirTag wallet offers a traditional design with a built-in AirTag holder.
Show more
If you’re in the market for an AirTag wallet, you already know the marketplace is extensive. Here are a few things to keep in mind while you search that may help you narrow down your options and find the right fit for you.
Capacity: Consider what you want to store in your wallet. How many cards? Cash? Are there any other incidentals like your work badge or transit passes? We’d recommend finding out your absolute bare minimum need for card capacity (maybe that’s three cards or maybe it’s ten) and factoring that into your search. You’ll easily be able to cut a huge chunk of options out based on capacity alone.
Security: Many AirTag wallets provide RFID-blocking, which means that they block card scanners that can steal your information. If this is a top priority for you, you can drop many of the cheaper options, which often don’t offer these capabilities.
Cost: The price of AirTag wallets can vary considerably, and like with any purchase, establishing a budget is key. While we’ve included a budget, midrange, and expensive pick on this list, you need to decide for yourself what your magic number is. Keep in mind that your wallet is often an everyday essential that lasts for a number of years, so it may be worth it to set a higher budget. Plus, many quality traditional leather wallets also cost over $100.
Your needs: You’re here because you think you want an AirTag wallet. But maybe you’re looking for a Bluetooth tracker card to upgrade your favorite traditional wallet. Or maybe you just want an AirTag card insert to do the same. Think about what you need right now and what makes the most sense for your lifestyle — it may not be what you initially thought.
Show more
We combined our expert research and analysis, hands-on testing for select products, and our extensive testing and knowledge of alternative Bluetooth trackers to curate this list. We consider factors like lifestyle, construction, durability, tracking capabilities, and function while compiling these top picks and made sure to provide options for those who aren’t sure if an AirTag wallet is right for them. If we think a new product is worth your money, we’ll update this list or write a spotlight review highlighting its best qualities and specifying who it’s for.
Plus, we are routinely scouring the web for new products that are worthy of testing, following news on product rollouts and upgrades, and sourcing the best deals.
Show more
An AirTag is a type of tracking device. You can affix it to an object, and the AirTag will then provide tracking for that item. Simply log into the Find My app to track your device and see an exact location. You can also ping your phone so it sounds like an alert, helping you to find it that much easier.
Show more
Yes, AirTags can be used to track anything like keys, bags, purses, wallets, and more. There’s lots of accessories on the market to seamlessly configure an AirTag to your needs, and likewise, many AirTag wallets are specially designed to hold an AirTag to upgrade your traditional wallet with location-tracking capabilities.
Also: The best AirTag accessories of 2024
Show more
While this may not be the conventional route, if you’re interested in tinkering and building your own custom AirTag setup, you actually can. ZDNET’s Adrian Kingsley-Hughes did himself in April 2023, and wrote up his own step-by-step guide that you can follow.
Show more
AirTags will not interfere with your debit or credit cards and they will not cause demagnetization. AirTags operate using Apple’s Find My technology and a bluetooth connection, and the frequency does not disturb your cards.
Show more
There’s a few different types of products and configurations that will help you track your wallet, no matter where you leave it. Here are the most common:
A wallet specially designed to house an AirTag tracker, either by design or with added attachments.
A Bluetooth tracker card that uses Apple’s Find My technology that can slide into an existing wallet like any other card.
A tracker card designed to hold an AirTag that can slide into an existing wallet like a card.
Wallets that are already enabled with their own Bluetooth tracking systems and don’t require an AirTag (these often use Apple’s Find My technology).
MagSafe wallet options are enabled with Bluetooth tracking (which also likely uses Apple’s Find My technology).
Show more
According to the Department of Homeland Security, RFID refers to Radio Frequency Identification (RFID) technology. This technology uses radio waves to identify people or objects.
There are devices that can read information contained in a wireless device or “tag” from a distance without making any physical contact or requiring a line of sight — for example, the chips in credit or debit cards. Wallets which feature RFID-blocking technology prevent these devices from accessing information stored in protected chips, which can help prevent your bank information from being compromised.
Show more
The beauty of an AirTag is that it can be tracked from anywhere. Your AirTag just needs to be within range of a Bluetooth connection while near an Apple device. It does not even have to be your own phone, so as long as you are close to an Apple device, you can use AirTag technology to locate your wallet.
Show more
If you’re an iPhone user, there’s a few ways to recognize if an AirTag is tracking you. Your phone may alert you of an “AirTag found moving with you.” This will happen if these things are true:
The AirTag is separated from its owner.
Your iPhone is awake.
Also: How to find out if an AirTag is tracking you
Show more
More AirTag wallets to consider
If you haven’t found an AirTag wallet that feels right for you, here are a few alternatives that may better suit your needs — it may not be a traditional wallet that you’re looking for at all.
Need the AirTag first?
ZDNET sourced external industry experts for this article. Here is more information about them:
Liz Ho: An avid camper and outdoor enthusiast who runs the Camping Guidance YouTube channel, which has 16.5K subscribers.
Ricky Allen: An interior designer and marketing director of Ever Wallpaper, which sells eco-friendly wallpaper murals.
Show more
Other AirTag accessories to explore
From keychains to mounts and more, there are many AirTag accessories to choose from. Here are a few others we’ve tested.
WASHINGTON, DC—Over the course of a nearly three-hour committee hearing Wednesday, the nominee to lead NASA for the Trump administration faced difficult questions from US senators who sought commitments to specific projects.
However, maneuvering like a pilot with more than 7,000 hours in jets and ex-military aircraft, entrepreneur and private astronaut Jared Isaacman dodged most of their questions and would not be pinned down. His basic message to members of the Senate Committee on Commerce, Science, and Transportation was that NASA is an exceptional agency that does the impossible, but that it also faces some challenges. NASA, he said, receives an “extraordinary” budget, and he vowed to put taxpayer dollars to efficient use in exploring the universe and retaining the nation’s lead on geopolitical competitors in space.
“I have lived the American dream, and I owe this nation a great debt,” said Isaacman, who founded his first business at 16 in his parents’ basement and would go on to found an online payments company, Shift4, that would make him a billionaire. Isaacman is also an avid pilot who self-funded and led two private missions to orbit on Crew Dragon. Leading NASA would be “the privilege of a lifetime,” he said.
The hearing took place in the Russell Senate Office building next to the US Capitol on Wednesday morning, in an expansive room with marbled columns and three large chandeliers. There was plenty of spaceflight royalty on hand, including the four astronauts who will fly on the Artemis II mission, as well as the six private citizens who flew with Isaacman on his two Dragon missions.
“This may be the most badass assemblage we’ve had at a Senate hearing,” said US Sen. Ted Cruz, R-Texas, chair of the committee, commenting on the astronauts in the room.
Committed to staying at the Moon?
However, when the meeting got down to brass tacks, there were sharp questions for Isaacman.
Cruz opened the hearing by stating his priorities for NASA clearly and explicitly: He is most focused on ensuring the United States does not cede any of its preeminence to China in space, and this starts with low-Earth orbit and the Moon.
“Make no mistake, the Chinese Communist Party has been explicit in its desire to dominate space, putting a fully functional space station in low-Earth orbit and robotic rovers on the far side of the Moon,” he said. “We are not headed for the next space race; it is already here.”
Cruz wanted Isaacman to commit to not just flying human missions to the Moon, but also to a sustained presence on the surface or in cislunar space.
In response, Isaacman said he would see that NASA returns humans to the Moon as quickly as possible, beating China in the process. This includes flying Artemis II around the Moon in 2026, and then landing the Artemis III mission later this decade.
The disagreement came over what to do after this. Isaacman, echoing the Trump administration, said the agency should also press onward, sending humans to Mars as soon as possible. Cruz, however, wanted Isaacman to say NASA would establish a sustained presence at the Moon. The committee has written authorizing legislation to mandate this, Cruz reminded Isaacman.
“If that’s the law, then I am committed to it,” Isaacman said.
NASA astronauts Reid Wiseman, left, Victor Glover, Christina Koch, and CSA (Canadian Space Agency) astronaut Jeremy Hansen watch as Jared Isaacman testifies on Wednesday.
Credit:
NASA/Bill Ingalls
NASA astronauts Reid Wiseman, left, Victor Glover, Christina Koch, and CSA (Canadian Space Agency) astronaut Jeremy Hansen watch as Jared Isaacman testifies on Wednesday.
Credit:
NASA/Bill Ingalls
Cruz also sought Isaacman’s commitment to flying the International Space Station through at least 2030, which is the space agency’s current date for retiring the orbital laboratory. Isaacman said that seemed reasonable and added that NASA should squeeze every possible bit of research out of it until then. However, when Cruz pressed Isaacman about the Lunar Gateway, a space station NASA is developing to fly in an elliptical orbit around the Moon, Isaacman would not be drawn in. He replied that he would work with Congress and space agency officials to determine which programs are working and which ones are not.
The Gateway is a program championed by Cruz since it is managed by Johnson Space Center in Texas. Parochial interests aside, a lot of space community stakeholders question the value of the Gateway to NASA’s exploration plans.
Ten centers and the future of SLS
One of the most tense interactions came between Isaacman and Sen. Maria Cantwell, D-Wash., who wanted commitments from Isaacman that he would not close any of NASA’s 10 field centers, and also that the space agency would fly the Artemis II and Artemis III missions on the Space Launch System rocket.
Regarding field centers, there has been discussion about making the space agency more efficient by closing some of them. This is a politically sensitive topic, and naturally, politicians from states where those centers are located are protective of them. At the same time, there is a general recognition that it would be more cost-effective for NASA to consolidate its operations as part of modernization.
Isaacman did not answer Cantwell’s question about field centers directly. Rather, he said he had not been fully briefed on the administration’s plans for NASA’s structure. “Senator, there’s only so much I can be briefed on in advance of a hearing,” he said. In response to further prodding, Isaacman said, “I fully expect to roll up my sleeves” when it came to ideas to restructure NASA.
Cantwell and other Senators pressed Isaacman on plans to use NASA’s Space Launch System rocket as part of the overall plan to get astronauts to the lunar surface. Isaacman sounded as if he were on board with flying the Artemis II as envisioned—no surprise, then, that this crew was in the audience—and said he wanted to get a crew of Artemis III to the lunar surface as quickly as possible. But he questioned why it has taken NASA so long, and at such great expense, to get its deep space human exploration plans moving.
He noted, correctly, that presidential administrations dating back to 1989 have been releasing plans for sending humans to the Moon or Mars, and that significantly more than $100 billion has been spent on various projects over nearly four decades. For all of that, Isaacman and his private Polaris Dawn crewmates remain the humans to have flown the farthest from Earth since the Apollo Program. They did so last year.
“Why is it taking us so long, and why is it costing us so much to go to the Moon?” he asked.
In one notable exchange, Isaacman said NASA’s current architecture for the Artemis lunar plans, based on the SLS rocket and Orion spacecraft, is probably not the ideal “long-term” solution to NASA’s deep space transportation plans. The smart reading of this is that Isaacman may be willing to fly the Artemis II and Artemis III missions as conceived, given that much of the hardware is already built. But everything that comes after this, including SLS rocket upgrades and the Lunar Gateway, could be on the chopping block. Ars wrote more about why this is a reasonable path forward last September.
Untangling a relationship with SpaceX
Some of the most intelligent questions came from US Sen. Andy Kim, D-New Jersey. During his time allotment, Kim also pressed Isaacman on the question of a sustained presence on the Moon. Isaacman responded that it was critical for NASA to get astronauts on the Moon, along with robotic missions, to determine the “economic, scientific, and national security value” of the Moon. With this information, he said, NASA will be better positioned to determine whether and why it should have an enduring presence on the Moon.
If this were so, Kim subsequently asked what the economic, scientific, and national security value of sending humans to Mars was. Not responding directly to this question, Isaacman reiterated that NASA should do both Moon and Mars exploration in parallel. NASA will need to become much more efficient to afford that, and some of the US Senators appeared skeptical. But Isaacman seems to truly believe this and wants to take a stab at making NASA more cost-effective and “mission focused.”
Throughout the hearing, Isaacman appeared to win the approval of various senators with his repeated remarks that he was committed to NASA’s science programs and that he was eager to help NASA uphold its reputation for making the impossible possible. He also said it is a “fundamental” obligation of the space agency to inspire the next generation of scientists.
A challenging moment came during questioning from Sen. Edward Markey, D-Mass., who expressed his concern about Isaacman’s relationship to SpaceX founder Elon Musk. Isaacman was previously an investor in SpaceX and has paid for two Dragon missions. In a letter written in March, Isaacman explained how he would disentangle his “actual and apparent” conflicts of interest with SpaceX.
However, Markey wanted to know if Isaacman would be pulling levers at NASA for Musk, and for the financial benefit of SpaceX. Markey pressed multiple times on whether Musk was in the room at Mar-A-Lago late last year when Trump offered Isaacman the position of NASA administrator. Isaacman declined to say, reiterating multiple times that his meeting was with Trump, not anyone else. Asked if he had discussed his plans for NASA with Musk, Isaacman said, “I have not.”
Earlier in the hearing, Isaacman sought to make clear that he was not beholden to Musk in any way.
“My loyalty is to this nation, the space agency, and its world-changing mission,” Isaacman said. Yes, he acknowledged he would talk to contractors for the space agency. It is important to draw on a broad range of perspectives, Isaacman said. But he wanted to make this clear: NASA works for the nation, and the contractors, he added, “work for us.”
A full committee vote on Isaacman is expected later this month after April 15, and if successful, the nomination would pass to the full Senate. Isaacman could be confirmed late this month or in May.
With Nintendo's April 2 Direct showcase over and done, we now know more about the Switch 2 than ever before, including its release date (June 5) and price ($450!). We're also keeping a close eye on how the recently-announced tariffs in the United States are affecting Nintendo's launch plans — the company has already paused pre-orders in the US that were set to take place on April 9.
Thanks to Nintendo's January 16 teaser trailer, we've long known what the Switch 2 will look like, and that it will arrive in 2025 — over eight years after its predecessor. The Nintendo Switch 2 looks very similar to the original system, albeit with a larger display, magnetic Joy-Cons and a sleeker design. The company also reaffirmed that the upcoming console will be backwards compatible with Switch games.
Looking to catch up on everything we know about the Switch 2? Read on.
What are the key new features of the Nintendo Switch 2?
Display and dock
The Nintendo Switch 2 has a 1080p 7.9-inch display with a 120Hz refresh and HDR compatibility. It also supports up to 4K output at 120Hz (with variable refresh rates) when docked.
Joy-Con and Pro Controller
The new Joy-Con 2 for Nintendo Switch 2 attaches magnetically with a button to release them. They have larger SL and SR buttons (the ones hidden when the Joy-Con are attached) to make playing multiplayer games on individual Joy-Cons more feasible. These buttons are also made of steel and attach to the Switch 2's magnets.
The analog sticks are no longer tiny nubs, and are closer in size to the ones you might find on a PS5 or Xbox controller. However, they do not have Hall effect sensors, the use of which can minimize the risk of the "stick drift" issue that plagued the original Joy-Cons.
The Joy-Con 2 have what Nintendo is calling “HD Rumble 2” built in, which seems to be a refinement of the original (and still very good) vibration function. Finally, each Joy-Con 2 has an optical sensor that allows you to use it as a mouse, and a C button, which we’ll get to in a minute.
Nintendo
The new Pro Controller for the Nintendo Switch 2 comes with everything you’d expect based on the refreshed Joy-Con 2: Namely HD Rumble 2 and the C button. There are also remappable GL/GR buttons around the back and a standard audio jack for connecting a headset.
Both options come with Amiibo support built in. The Switch 2 comes with two Joy-Con 2, as you’d expect, and additional pairs are priced at $90. The Pro Controller will cost $80.
The C button stands for “chat”
The new C button is a dedicated way to enter a chat session with friends. As well as the standard features you’d expect from voice chat, Nintendo has built a Discord-like video-sharing feature, which lets you show your gameplay to others and see other people’s screens while you’re playing.
The GameChat function works with a mic built into the console, though headsets are also supported. Nintendo will also sell a $50 camera that plugs into the USB-C port on the top of the console, which will allow you to stream your face along with your game.
Nintendo
GameCube support
The Switch 2 will work with GameCube games via the Nintendo Switch Online Expansion Pack. The three titles available at launch will be F-Zero GX, The Legend of Zelda: The Wind Waker and Soul Calibur II.
Improved specs
Perhaps the biggest feature, though, is one you can’t see: The Switch 2 has an all-new processor and GPU and significantly more storage (256GB vs 64GB in the Switch OLED), along with support for faster and more capacious microSD cards. This will obviously lead to better first-party games and upgraded Switch experiences, but more importantly it will mean multi-platform games that had to skip the original, underpowered Switch will be able to be ported over. With games like Cyberpunk 2077 coming to the Switch 2 at launch, it seems like far more ports will be feasible than on the original console.
Another major improvement comes in the form of networking — the Switch 2 supports Wi-Fi 6, which will improve the original's often glacial download speeds. Similarly, the new dock sports an Ethernet port for a rock-solid connection.
What's it like to play?
Nintendo gave the press a chance to check out the Switch 2 and a few of the games it highlighted. You can read our impressions of Mario Kart Worldas well as more general thoughts on how the Switch 2 feels in person. Right off the bat, Mario Kart World feels like an outstanding launch title, with more depth and better visuals than we've ever seen from a Mario Kart game. The core formula is intact, but there are a lot of new features and play styles to dig into.
As for the Switch 2 hardware, it feels like a major step forward from the original Switch — it has grown up a lot and feels far more polished and refined. And while it may not have an OLED screen, the 7.9-inch LED display Nintendo did use is outstanding.
When will the Nintendo Switch 2 come out?
June 5, 2025.
When do pre-orders open for Nintendo Switch 2?
Pre-orders were originally set to begin on April 9, but on April 4 Nintendo announced it was suspending pre-order plans in the US "indefinitely" in reaction to the original slate of tariffs imposed by President Trump. However, Nintendo says that the release date for the console is unchanged.
How much will the Nintendo Switch 2 cost?
On April 2, Nintendo announced that the Switch 2 would cost $450 standalone, or $500 with a bundled digital copy of the new Mario Kart game, Mario Kart World. That’s significantly more expensive than the original Switch, which launched at $300, and has remained at that price ever since. It’s also more expensive than the entry-level current-gen consoles from Sony and Microsoft.
The comparison looks a little better up against Valve’s Steam Deck, which costs $400 for the LCD model or $550 for the basic OLED model. The Steam Deck is more affordable than most PC handhelds.
Within 48 hours of the pricing announcement, however, Nintendo delayed the original pre-order schedule "in order to assess the potential impact of tariffs and evolving market conditions." With heavy US tariffs levied on Nintendo's Asian production centers, the implication was that would have little choice but to pass at least some of those costs on to American consumers. But Trump's April 9 tariff pivot — reverting to 10 percent across the board for 90 days for all countries except China, which now gets a triple-digit hit — further muddied the waters. That's because it's unclear how much, if any, of the components and assembly of US-bound Switch 2 consoles will originate in China.
How long will the Switch 2's battery last?
Nintendo says that the Switch 2 will last between 2 and 6.5 hours on a single charge. This is similar to the original Switch, which was rated for 2.5-6.5 hours, though later revisions upped that figure significantly. The company cautions that "this time is an estimate … battery life will depend on the games you play and usage conditions."
When can I try the Nintendo Switch 2?
Nintendo is planning a worldwide roadshow to let gamers go hands-on with the console. These events started in New York and Paris on April 4-6, with others taking place throughout the following two months.
Ticket registration for many of the Nintendo Switch 2 Experience events has closed, but waitlists are available. However, given that Nintendo is taking a first come, first served approach, your chances of attending the roadshow if you don't already have a ticket (or unless you signed up for the waitlist immediately) seem very small at this point.
The full list of Nintendo Switch 2 Experience dates is as follows:
North America:
New York, April 4-6, 2025
Los Angeles, April 11-13, 2025
Dallas, April 25-27, 2025
Toronto, April 25-27, 2025
Europe:
Paris, April 4-6, 2025
London, April 11-13, 2025
Milan, April 25-27, 2025
Berlin, April 25-27, 2025
Madrid, May 9-11, 2025
Amsterdam, May 9-11, 2025
Oceania:
Melbourne, May 10-11, 2025
Asia:
Tokyo (Makuhari), April 26-27, 2025
Seoul, May 31-June 1, 2025
Hong Kong, To be announced
Taipei, To be announced
Is the Nintendo Switch 2 backwards compatible?
Nintendo confirmed in November 2024 that the Nintendo Switch 2 will be backwards compatible. It will also feature access to Nintendo Online, so users will be able to play all of those old retro titles.
In the initial Nintendo Switch 2 press release, Nintendo reiterated that physical and digital Switch games will work on the new system. However, it noted that "certain Nintendo Switch games may not be supported on or fully compatible with Nintendo Switch 2." We now have more information on which games are working thanks to a guide Nintendo put together.
Nintendo's software compatibility chart on April 2, 2025.
Nintendo
As of April 1, the vast majority of Switch 2 games are marked as compatible, but many popular games are said to have some issues. Nintendo has marked most of these as “under investigation,” suggesting a fix may be on the way in time for launch. Nintendo says it is manually testing every Switch game for compatibility.
Only one title is marked as incompatible: Nintendo Labo Toy-Con 04: VR Kit. This was part of Nintendo’s Labo range of games that worked with Cardboard accessories, and is incompatible as the Switch 2 doesn’t physically fit in the VR headset.
Will old games be enhanced in any way?
The original Switch has, to put it mildly, struggled to run some of the late-generation software that's come down the pike. Could these games be enhanced to take advantage of the increased horsepower of the Switch 2? That’s unclear. In an “Ask the Developer” page on its site, Nintendo says that old games are working through a real-time translation system, which sounds similar to how Rosetta allows modern Macs to run old software. With that said, the new and old Switch are much closer in hardware than that.
In the same developer interview, it’s said that in its testing of old games for compatibility, there were some occasions “where loading times became faster, or game performance became more stable,” but we don’t have any detail beyond that.
As well as backward compatibility, Nintendo is also offering up premium "Switch 2 Edition" upgrades for a select few games. These have improved resolutions and frame rates, and (in some titles) some bonus features and content. What exactly that means will vary, but Nintendo did confirm that Metroid Prime 4: Beyond will have 4K/60 fps and 1080p/120 fps options on the Switch 2.
Games confirmed to have a Switch 2 Edition to date include:
Super Mario Party Jamboree
The Legend of Zelda: Breath of the Wild
The Legend of Zelda: Tears of the Kingdom
Kirby and the Forgotten Land
Pokémon Legends: Z-A
Metroid Prime 4: Beyond
In the case of the two mainline Zelda games, those upgrades will be included with a Switch Online + Expansion Pack membership or cost $10 each. Upgrades for the other games are also paid, but Nintendo has not said how much those will cost.
What will the cartridges taste like?
You may remember that the original Switch cartridges taste absolutely awful. This was on purpose, to discourage folks from putting the games in their mouth.
Similarly, the Switch 2 carts are also said to taste terrible, in large part to stop children or pets from accidentally ingesting them. It is not yet clear what substance Nintendo has coated the Switch 2 cartridges in to make them taste foul, but Nintendo's Takuhiro Dohta advised against licking them all the same.
"We don't want anybody to be at risk of any unwanted consumption," he told GameSpot. "We have indeed made it so that if it enters your mouth, you'll spit it out."
What are the launch games?
We expect this list to change substantially over the coming weeks and months, but the titles currently confirmed to be available on day one follow:
Arcade Archives 2 Ridge Racer
Bravely Default: Flying Fairy HD Remaster
Cyberpunk 2077
Deltarune
Fast Fusion
Fortnite
Hitman: World of Assassination
Hogwarts Legacy
Kunitsu-Gami: Path of the Goddess
Mario Kart World
Nintendo Switch 2 Welcome Tour
Puyo Puyo Tetris 2S
Split Fiction
Street Fighter 6
Survival Kids
Yakuza 0: Director's Cut
That's everything we know about the Nintendo Switch 2 today. We'll update this article with any information we can gather directly from sources. Any changes made to the article after its initial publishing will be listed below.
Update, January 24, 2025, 12:36 PM ET: This story has been updated to include speculation about price, potential game enhancements and the taste of cartridges.
Update, February 5, 2025, 9:30AM ET: This story has been updated to note the time when the April 2 Nintendo Direct starts.
Update, February 24, 2025, 12:30 PM ET: This story has been updated to include speculation about storage and information about the new microSD Express standard.
Update, March 6, 2025, 2:30PM ET: This story has been updated to note recent FCC filings to indicate the presence of Wi-Fi 6 and NFC support.
Update, March 14, 2025, 12:15PM ET: This story has been updated to include pricing and sales speculation from analysts.
Update, March 20, 2025, 12:45PM ET: Updated to include a note about the Seattle Mariners new jersey patches featuring Nintendo and the Switch 2.
Update, March 27, 2025, 3:50PM ET: Updated to add details about the C button, Nintendo Today! and Virtual Game Cards.
Update, April 1, 2025, 10:16AM ET: Updated to add link to livestream and details on its length.
Update, April 2, 2025, 4:40PM ET: Added details from the April 2 Nintendo Direct event, including specifications, price, release date and launch titles.
Update, April 4, 2025, 11:17AM ET: Added details about pre-orders being put on hold.
Update, April 7, 2025, 1:56PM ET: Added details about the Switch 2 cartridges still (reportedly) tasting terrible, the lack of Hall effect sensors in the controllers and game upgrade pricing.
Update, April 9, 2025, 5:41PM ET: Added context on how the changing rules on the Trump tariffs may or may not affect Switch 2 pricing in the US.
Jeff Dunn contributed to this report.
This article originally appeared on Engadget at https://www.engadget.com/gaming/nintendo/nintendo-switch-2-updates-release-date-price-new-games-and-everything-else-you-need-to-know-175623042.html?src=rss
Self-driving cars are slowly becoming less sci-fi and more real-world as companies like Waymo, the driverless arm of Google’s parent company Alphabet, expand into more cities — and countries.
On Wednesday, Waymo said it’ll begin driving its vehicles on Tokyo’s streets, making this the company’s first international location. Waymo announced the expansion in December, which is happening in partnership with Japanese taxi service Nihon Kotsu and taxi app Go. Now, the US-based company is finally ready to hit the road abroad.
Waymo says trained Nihon Kotsu drivers will manually drive its vehicles across seven Tokyo wards, including Minato, Shinjuku, Shibuya, Chiyoda, Chūō, Shinagawa and Kōtō. This will allow engineers to test and adapt Waymo’s autonomous driving tech to local road features and traffic.
“In Tokyo, we are abiding by the same steadfast principles that guide us in the US — commitment to safety, dedication to earning trust in communities where we operate, and collaboration with local officials and community groups here in Tokyo,” Nicole Gavel, Waymo’s head of business development and strategic partnerships, said in a statement.
Waymo currently operates fully autonomous rides for the general public in San Francisco, Phoenix, Los Angeles and Austin, Texas. In January, the company announced it would also begin testing with manually driven vehicles in 10 new cities this year, starting with Las Vegas and San Diego. It’s also slated to expand into Atlanta later this year through a partnership with Uber, and will begin offering rides in Washington, DC, through its Waymo One app starting in 2026.
Autonomous rides take place aboard the all-electric Jaguar I-Pace, but in October, Waymo announced it’s partnering with Hyundai to bring the next generation of its technology into Ioniq 5 SUVs. In the years to come, riders will be able to summon those all-electric, autonomous vehicles using the Waymo One app.
Waymo says it provides more than 200,000 paid trips each week. I’ve hailed several rides myself in San Francisco, and as off-putting as it can seem at first (especially to see a steering wheel turn by itself), I quickly adjusted and it soon felt like an ordinary ride.
That’s not to say there hasn’t been pushback as Waymo rolls out to more cities. The company’s vehicles have been involved in a handful of high-profile collisions, including one with a biker in San Francisco, and another with a towed pickup truck in Phoenix. (It recalled and updated its software to address the issue.) In response to concerns, Waymo says that across 14.8 million miles, its autonomous Waymo Driver technology “was up to 3.5x better in avoiding crashes that cause injuries and 2x better in avoiding police-reported crashes than human drivers in SF and Phoenix.” It also released a data hub last year, detailing efforts to improve road safety using its technology.
As Waymo continues to expand and develop its self-driving tech, here’s how and where to summon the robotaxi if you happen to be in one of the few cities where the company currently operates its fleet.
Watch this: Testing Waymo’s Safe Exit Feature in a Self-Driving Taxi
Hailing a ride in Phoenix
Phoenix was the first city to open up fully autonomous Waymo rides to the public, in 2020. To hail a ride, download the Waymo One app on iOS or Android. The service operates 24 hours a day, seven days a week.
You can also use the Uber app to summon one of Waymo’s vehicles in Phoenix. When you request an UberX, Uber Green, Uber Comfort or Uber Comfort Electric ride, you’ll have the choice to confirm a Waymo ride, if you’re matched.
In addition to hailing a ride, you may also have your Uber Eats meal delivered by an autonomous car. When placing an order in the Phoenix area, you might get a note that “autonomous vehicles may deliver your order.” When the Waymo car arrives, take your phone with you to pop open the trunk and grab your delivery. You can opt out of this during checkout if you’d rather have a human deliver your food.
Hailing a ride in San Francisco
San Francisco followed suit after Phoenix, rolling out fully autonomous rides in late 2022. It scrapped the waiting list in June, so now anyone can download the Waymo One app to ride anytime. The service also operates 24 hours a day, seven days a week. There’s currently no Uber partnership in San Francisco.
In August, Waymo expanded its ride-hailing service into the San Francisco Peninsula, adding 10 square miles and venturing into Daly City, Broadmoor and Colma. And in March, it expanded again so that select Waymo One riders who live in Mountain View, Los Altos, Palo Alto and parts of Sunnyvale have the option to hail a robotaxi there, too. The company says it plans to expand to more riders there over time.
Hailing a ride in Los Angeles
In November, Waymo scrapped its waitlist for Los Angeles and began welcoming all public riders via the Waymo One app. Now any interested passengers can hop in the robotaxis 24/7 and ride across nearly 90 square miles of LA county, including Santa Monica, Beverly Hills, Inglewood and Downtown LA — with plans to expand into more areas soon.
Hailing a ride in Austin
In Austin, the only way to hail a Waymo ride is through Uber — no Waymo One app here. By requesting an UberX, Uber Green, Uber Comfort or Uber Comfort Electric, you could be matched with a Waymo vehicle — and you won’t be upcharged. If you’d rather not take a driverless ride, you’ll have the option to switch to a standard one. On the other hand, if you want to boost your chances of being matched to a self-driving car, you can go to Account > Settings > Autonomous vehicles, then hit the toggle next to Get more Waymo rides.
Unlock the door, pop open the trunk and start the ride from the Uber app. You’ll still be asked to rate your ride at the end — but you won’t be asked to tip.
If there are any issues, riders can access human support 24/7 via both the Uber app and from inside the Waymo vehicle (there are screens in the front and back that let you quickly summon customer support).
As part of the Uber partnership, Uber will manage tasks like vehicle cleaning and repair, while “Waymo will continue to be responsible for the testing and operation of the Waymo Driver, including roadside assistance and certain rider support functions,” the companies said. The collaboration should make autonomous rides accessible to more people, who now won’t have to download a separate app to hitch a ride in a robotaxi.
Waymo currently operates across 37 square miles of the city, with plans for future expansions.
Coming soon: Atlanta, Washington, DC, and Miami
In Atlanta, public riders can catch a driverless ride via the Uber app starting in early 2025. Waymo employees are currently the only people in Atlanta who can hail the fully autonomous vehicles.
Members of the public in Atlanta can join an interest list and be notified when it’s time to ride.
Meanwhile, Waymo is planning to begin offering rides in Washington, DC, in 2026. The company returned to the nation’s capital in January to test its autonomous driving tech. In late March, it said it was bringing more vehicles to the city and working to scale its service throughout the year. In a blog post, Waymo said it’ll “continue to work closely with policymakers to formalize the regulations needed to operate without a human behind the wheel in the District.”
You’ll also have to wait until 2026 to ride a Waymo in Miami. The company has been conducting weather testing in the lead-up to that eventual rollout, noting in a blog post, “Our previous road trips to the Sunshine State’s challenging rainy conditions have been invaluable in advancing our autonomous driving capabilities.”
Waymo will collaborate with Moove, a fintech company that offers vehicle financing, first in Phoenix, where Moove will manage the robotaxi’s fleet operations, facilities and charging infrastructure. In both Phoenix and then Miami, “Waymo will continue to offer our service through the Waymo One app, and remain responsible for validation and operation of the Waymo Driver,” the company said in a blog post.
The road ahead: Future vehicles
In August, Waymo unveiled the sixth generation of its self-driving technology, which aims to expand the capabilities of its driverless fleet. Smarter sensors are meant to help the cars better navigate in extreme weather, Waymo said. The sixth-gen Driver will come aboard the all-electric Zeekr vehicle, which features a flat floor, more head- and legroom, adjustable seats and a removable steering wheel and pedals. The updated tech is still being tested, and the company says it’ll be available to riders soon.
In October, Waymo also announced a partnership with Hyundai to integrate its sixth-generation Driver into the all-electric Ioniq 5 SUV, which, according to a blog post, “will be added to the Waymo One fleet over time.” The companies added that they “plan to produce a fleet of Ioniq 5s equipped with Waymo’s technology in significant volume over multiple years to support Waymo One’s growing scale.” Testing with these vehicles will start by late 2025 and become available “in the years to follow.”
Currently, everyday riders can’t take a Waymo vehicle on highways and freeways, but that may soon change. The company is testing fully autonomous rides on freeways in Phoenix, as seen in an exclusive video shared with CNET in May. It’s also now making fully autonomous freeway rides available to Waymo employees in Los Angeles.
Waymo is working to expand its autonomous driving tech into trucking as well, but it said last year that it’s scaling back those efforts for the time being, to focus on ride-hailing with Waymo One. It noted, “Our ongoing investment in advancing Waymo Driver capabilities, especially on freeways, will directly translate to trucking and benefit its development efforts.”
The sci-fi fairytale of de-extinction may become a reality. Colossal Laboratories and Biosciences announced they have revived dire wolves from extinction, unveiling three canines it has genetically engineered to resemble the long-dead species.
On Tuesday, Colossal revealed that two genetically identical male “dire wolf” puppies named Romulus and Remus were born on Oct. 1 last year, with their sister Khaleesi following a few months later on Jan. 30. (Sansa or Arya may technically be more appropriate Game of Thrones-inspired names for a dire wolf, or even Lady or Nymeria, but you can’t deny that “Khaleesi” is immediately recognizable.)
SEE ALSO:
Scientists accidentally created the cutest mouse in the world
“They’re beautiful, they’re cute, they’re incredible,” said Ben Lamm, co-founder and CEO of Colossal Biosciences told Mashable, sounding every bit the enthusiastic new dog parent. “We took a 13,000-year-old tooth and a 72,000-year-old skull, and made puppies.”
Created from edited grey wolf genomes, Colossal states that these three canines are proof that a “standardized toolkit for de-extinction” is possible.
How did Colossal create its dire wolves?
Colossal’s dire wolf puppies Romulus and Remus at 15 days old. Credit: Colossal Laboratories and Biosciences
Directly using dire wolf cells to clone a new animal isn’t currently possible, as such material simply doesn’t exist. Instead, Colossal used ancient DNA and comparative genomics, following a methodology similar to the one in its woolly mouse project. Revealed in March, Colossal’s woolly mice were created by editing the genome of normal mice to display physical traits similar to those found in woolly mammoths — most significantly their long, fluffy fur.
For its dire wolf project, Colossal began with the aforementioned 13,000-year-old tooth from Ohio and 72,000-year-old inner ear bone from Idaho. Sampling these specimens and building on previous findings, Colossal’s team extracted and sequenced dire wolf DNA, assembled genomes, then compared the results to that of living canines. This enabled them to pinpoint where dire wolves’ genomes differ from that of other species. In doing so, they discovered that grey wolves share 99.5 percent of their DNA with dire wolves, making them the extinct animal’s closest living relative.
Using this information, Colossal then employed a gene-editing method called CRISPR to edit the grey wolf’s genome to more closely resemble that of a dire wolf. To do this, they made 20 strategic edits across 14 genes, with 15 of the edits recreating extinct gene variants. The nuclei were then removed from the resultant cells and transferred into donor egg cells, creating embryos that Colossal transplanted into surrogates. Leave it to bake, and soon you have a dire wolf pup.
It wasn’t a simple process though, and just as in human fertilisation treatments, not every embryo resulted in a pregnancy. Colossal made multiple attempts in an effort to increase the odds of success, completing eight embryo transfers, involving between 30 to 45 specimens each.
“There were a couple of times where a pregnancy was established early on, but was resorbed, and in the end we ended up with these three animals that are healthy,” Colossal’s chief science officer Beth Shapiro told Mashable.
Creating dire wolves from blood
A Colossal Laboratories and Biosciences’ dire wolf at five months old. Credit: Colossal Laboratories and Biosciences
Colossal also announced that it had developed a new, less invasive cloning method to create its dire wolves, using cells never previously used in such procedures. Specifically, its scientists used endothelial progenitor cells (EPCs) gathered by drawing an animal’s blood, as opposed to using other cells retrieved via a skin biopsy.
EPCs aren’t undifferentiated stem cells, but they haven’t been entirely differentiated either, falling somewhere in between the two. Shapiro told Mashable that the cells Colossal’s new blood draw method extracted were of a higher quality than other cells previously used in their research, being healthy and easily grown.
“These progenitor cells will eventually become the cells that make up the lining of the blood vessels,” said Shapiro. “Because they’re not all the way down the path to having a very specific job as a cell, it’s easier to reprogram them. And we’ve also found that it’s easier to establish really healthy cell lines, so it’s great for biobanking.” (Biobanking is the collection and storage of biological samples for conservation or research purposes.)
Colossal also used the same method to clone four red wolves — one female and three males. A critically endangered species, it’s estimated there are fewer than 20 red wolves currently in the wild. Shapiro said that though they only conducted two embryo transfers for the red wolves, both resulted in successful pregnancies.
“Normally when you’re doing this, the reprogramming step is such that it takes many more tries than that to have a successful pregnancy from a cloned cell,” said Shapiro. “So we’re pretty confident that this is a really great technology that can be used for cloning animals in the future.”
“We’re pretty excited because we think that de-extinction and species preservation goes hand in hand,” said Lamm.
What makes a wolf a dire wolf?
Two-month-old dire wolf Khaleesi stretching open her mouth. Credit: Colossal Laboratories and Biosciences
Whether or not Colossal’s dire wolves are “real” dire wolves depends on how strictly one defines the species. Technically, Colossal doesn’t have any viable material straight from a dire wolf to work with. The company didn’t find this an insurmountable obstacle though, working with a grey wolf genome base as an acceptable substitute instead.
“We focus on what genes kind of drive — like, what makes a dire wolf a dire wolf, and then how do we engineer them into the grey wolf,” said Lamm “Because they are 99.5 percent the same genetically.”
Colossal discovered during its DNA sequencing that dire wolves would have had thick white coats. They were also around 25 percent larger than grey wolves, with wider heads and stronger jaws. Accordingly, Colossal’s wolves have a distinctly more muscular build than the leaner grey wolf. They’re also already over 80lbs (36kg) at six months old, which is a comparable weight to a fully grown grey wolf. As such, Lamm expects that the dire wolves may end up weighing around 140lbs (64kg) once they reach adulthood.
“When we talk about de-extinction, we’re talking about bringing back the core features, the core traits that resemble — that are — the species that used to be alive,” explained Shapiro. “There are many different ways that people can consider what a species is. But we’re talking about this functional trait. So we’re bringing back these functional [traits].
Mashable Light Speed
“We know that dire wolves were larger, more muscular, that they had this particularly interesting coat, thick coat. And so when we target the DNA that we’re going to edit… those 20 edits brought back these dire wolf phenotypes, and the animals that are born are our dire wolves.”
This isn’t Jurassic Park
A Colossal Laboratories and Biosciences’ dire wolf. Credit: Colossal Laboratories and Biosciences
Unsurprisingly, Colossal’s de-extinction efforts have repeatedly drawn comparisons with the sci-fi film Jurassic Park. This association is only likely to continue now that their first de-extinction project has made a larger, stronger version of a charismatic carnivore. However, Lamm notes that there are important differences between Jurassic Park and Colossal’s work, and that its real scientists have very different goals from the movie’s fictional ones.
“The three reasons why we built the dire wolves in the first place was: One, we were working with some Indigenous tribes here in the United States that actually asked us to bring back the dire wolves,” said Lamm. “They said it was akin to the Great Wolf in their culture, they said it has ancestral knowledge and it’s spiritually important to them. So I don’t think they did that in Jurassic Park.”
Specifically, Lamm noted that the Mandan, Hidatsa, and Arikara Nation (MHA Nation) was Colossal’s biggest partner on its dire wolf project. They also partnered with the Nez Perce Tribe, the Karankawa Tribe of Texas, INDIGENOUS LED, and the Wind River Tribal Buffalo Initiative.
“Number two, we were working on building technologies for conservation of wolves, and to specifically save the red wolves from the brink of extinction that it’s on,” Lamm continued, referring back to Colossal’s cloning of the endangered species. “Unless there was a subplot I missed in Jurassic Park, I don’t think that they were focused on the conservation aspects.
“And then the third thing is… we do want to inspire the next generation, get kids excited about the science and whatnot,” said Lamm. “Maybe Jurassic Park tried to do that in their ‘theme parks.’ But it’s definitely a very different set of goals.”
What happens to Colossal’s dire wolves now?
Colossal’s dire wolves won’t simply be released into the wild to repopulate. They’re currently being kept on an enclosed, protected preserve measuring over 2,000 acres, with potential plans to eventually move them to a larger such space on Indigenous land. The company further noted that the preserve is certified by the American Humane Society, and registered with the United States Department of Agriculture.
The wolves further have a 10-person team looking after them, and are constantly monitored with cameras and drones like a dire wolf Truman Show. This team manages factors down to the wolves’ social dynamics and moves them through feeding cycles.
“They have 24-hour veterinary care,” said Shapiro. “These animals can’t get a splinter without us knowing about it. So they’re not exactly living like wild wolves. They’re living the luxury style of life for wild wolves.”
One of Colossal’s dire wolf pups being handfed with a bottle at 15 days old. Credit: Colossal Laboratories and Biosciences
They’re also being hand-raised by humans. While Romulus, Remus, and Khalissi were initially cared for by their surrogate mothers, Colossal eventually separated them due to concern the dog parents weren’t up to the task of caring for their large dire wolf pups.
“The mother was getting a little overzealous, she was like over-cleaning them, moving them away,” said Lamm.
There were also issues with the surrogate mothers disturbing the puppies’ sleep by continually waking them to feed, as well as concerns regarding the amount of nutrition the mothers’ milk could realistically provide.
“[The dire wolf puppies] were larger than a standard grey wolf, and we wanted to make sure that they were getting enough food,” said Shapiro. “The surrogates were large hound domestic dogs, so… they might not have produced as much nutrition as a large wolf would have needed. So we were just being overly careful [by separating them].”
A three-month-old dire wolf holding a stick in the snow. Credit: Colossal Laboratories and Biosciences
There is the issue of how the dire wolves will learn to be dire wolves, considering that there aren’t any animals of the same species to teach them how. The wolves’ human parents can’t very well pass on skills such as how to take down a deer. However, Shapiro told Mashable that the dire wolves are already displaying innate behavior similar to that seen in wild wolves or those raised in captive or semi-captive environments.
“A lot of what we see is behavior that we predicted to see in these wolves,” said Shapiro. “That’s another thing that’s really great about this group as our first de-extinction project, because we can watch these animals sort of interact with their habitat and see how they learn from each other, what they pick up from each other and from other animals with which they interact in their habitat, and we will gradually make their experience in life more and more similar to the wild.
“But mostly what we’re focused on right now is making sure that they’re healthy. We want to make sure that they are growing the way they should grow. We want to keep tabs on this because we really want to understand the impact of the edits that we made on their growth patterns and their physiology and their behavior.”
This includes the dire wolves’ social dynamics. Though Khaleesi is currently being kept separate from her brothers, Shapiro told Mashable that she’s getting old enough to be introduced to them soon.
A dire wolf puppy at one month old. Credit: Colossal Laboratories and Biosciences
Colossal is also being cautious to avoid unintended dire wolf puppies. Having the dire wolves reproduce au naturel would no doubt be exciting in theory, and the company is indeed planning on creating three to five more of them to build the pack dynamics it wants. However, any such additions must be carefully planned and monitored.
As such, in addition to tracking the animals’ reproductive cycles and keeping them strategically separated, Colossal has also administered wolf birth control in the form of subdermal hormonal contraception.
“We really need to manage the population, study them, decide what the long term plan is with rewilding, [and] on what size area,” said Lamm. “Potentially on Indigenous land, on MHA Nation, for example. And then we’ll go from there.”
A three-month-old dire wolf in the snow. Credit: Colossal Laboratories and Biosciences
In the meantime, Romulus, Remus, and Khalissi are apparently living relatively pampered lives by wolf standards. Lamm told Mashable that the wolves initially ate colostrum and milk, first from their mothers and then synthetic versions. Now they’re eating a combination of horse, bison, and elk meat, as well as some dry food. Colossal is also planning to introduce carcasses soon to enable the dire wolves to figure out socialisation and behavior surrounding mealtimes.
While the dire wolves haven’t hunted for themselves yet, Lamm stated that their future diet would likely be a combination of hunted and supplied meals. Though technically, all of the meals would be supplied.
“It’s not like we’re going to let them out to compete with grey wolves,” said Shapiro. “They’re on a protected reserve, so if they are introduced to hunting it will have to be because we’ve put animals or the trainers or the handlers have put animals there.”
“They’re starting to exhibit hunting-like behaviors,” said Lamm. “So they’re starting to act more and more wolf-like by the minute.”
Though it may be beneficial for the dire wolves to learn to hunt from a development perspective, they’re unlikely to ever genuinely need the skill, living out their days under researchers’ watchful care. Colossal doesn’t have a definitive idea of its dire wolves’ life expectancies, however Lamm noted that regular wolves can live for 10 to 15 years.
It’s a fair estimate. Even so, exactly how long genetically engineered dire wolves live is something we’ll only find out through lengthy observation. These are the first ones, after all.
Americans are taking to the streets and showing up in US cities by the thousands to protest a plethora of President Donald Trump’s second-term policies, and to challenge the power that Trump has bestowed upon Elon Musk to disrupt the federal government. Public outcry and acts of civil disobedience have become more commonplace over the past several years, as citizens have challenged Trump’s threats to women’ s rights, travel bans from foreign nations and deportation policies, police brutality against Black Americans, and the US Supreme Court’s overturning of Roe v. Wade.
Now the protests, which include the Tesla Takedown movement and a nationwide mass-action called Hands Off, are targeting a hostile takeover of the government and broader threats to American freedom. (Trump has also seriously threatened longstanding free-trade agreements and single handedly taken down the US stock market in the process.)
This WIRED guide to safe protesting was originally written in 2020 during the nationwide outcry over police brutality, which overwhelmingly targets Black people like George Floyd, Breonna Taylor, and Tony McDade. We’ve now updated this guide to include advice specific to mass action, your rights and protections as a US citizen, protesting in or around government property, and laws around defacing property (like, say, a Tesla).
If you’re thinking about joining a protest near you, there are some crucial factors to consider: The US government is now allowing federal immigration agents to conduct sweeping raids, and ICE is reportedly detaining and deporting people in the US who are here on legal visas, including many students. And, while police brutality is an abstract concept for some, it’s a stark reality for others. There are other ways you can contribute if you don’t feel safe protesting or are otherwise unable to physically do so. You can also donate time to community groups, drop off supplies for protests, or contact local legislators. Now, more than ever, it’s crucial to put pressure on our legislators.
That being said, protesting is a right of all Americans under the First Amendment (more on that below). Before you head out, you should know that police across the country have acted with unnecessary force, including driving vehicles through crowds, partially blinding a photojournalist, and macing children. The list goes on and on.
If you still want to join in, we’ve gathered some advice, as well as a list of items you may want to bring with you. Be careful, and stay safe.
Table of Contents
What to Bring (and Not Bring) to a Protest
It’s smart to have supplies on hand for a day of protesting. We recommend the following. You probably have everything you need around the house, and if not, these items will likely be accessible at your local stores. We’ve included some links to online retailers for your reference.
A bag and/or backpack: You’ll need something small and durable. I, Louryn, use a cheap daypack from Walgreens for most supplies, and I also strap on a belt bag, which I use to hold the essential items I’d need if I were to lose my backpack. Use whatever you own that lets you keep your hands free. If you don’t have anything, we have a list of our favorite fanny packs. And while your bag should be big enough to hold all the supplies you need, be sure to avoid anything too bulky.
Water: There’s a good chance that your protest will include a march, so you need to pay attention to proper hydration. Carry drinking water. Bring the biggest bottle that you can fit in your bag. Water can also be used to clean wounds and flush the eyes of anyone who’s been hit with chemical gas or pepper spray.
A face mask or bandana: When we first wrote this guide in 2020, we advised wearing masks because of the Covid-19 pandemic. That’s less of a concern for many people now, but you still may want to take precautions and cover your face with a mask or bandana. Face coverings also have the benefit of shielding your identity from cameras and police surveillance. (We have a separate guide about digital privacy during protests.). Bring an extra mask if you have one.
A hat and/or sunglasses: Aside from shielding you from the sun during a long day of marching, hats and sunglasses can obscure your face from surveillance and protect your privacy. If you wear a hat, and you’re interested in further protecting your identity, keep the brim low.
Snacks: You are likely in for a long day. Pack lightweight, nutritious, protein-rich snacks. Jerky, energy bars, and nuts are all good picks.
Protest signs: If you want to carry a sign, there are some things to consider. Ensure that your slogan is in big, bold letters that can be easily read from far away. Short and punchy sayings are arguably better than a block of script. Poster board is flexible, but stiffer foam-core board is more durable. You can affix paint-stir sticks or other flat, wooden sticks to the sign using strong tape to create a handle. You might want to make extras to hand out to fellow protesters. Don’t litter—when you’re done with your sign, dispose of it properly, or donate it to another protester.
Suitable clothing: It’s a good idea to wear all black, both because that’s what the organizers of most solidarity protests suggest and because it helps you blend in with a crowd. Privacy experts also recommend that you cover any tattoos, if you can, and that you hide your hair if it’s dyed a distinctive color.
A change of clothes: If you’re protesting on a particularly hot day, you may want to have extra clothes. These can also come in handy if you’re exposed to substances that can hurt your skin or if you’re splashed with paint, gross road water, or other people’s sweat. I usually carry a pair of shorts, a tank top, and an extra pair of socks in my backpack.
Hand sanitizer: You might find yourself holding hands with a stranger, grabbing onto gunky street signs, or tripping and falling into a puddle. All these scenarios coupled with Covid-19 make hand sanitizer an essential thing to carry.
Good walking shoes: This is nonnegotiable. Wear closed-toe shoes that are broken in and good for walking long distances.
Your ID (maybe): If you’re detained, not having your ID on you might keep you stuck for longer. However, in some states, you might not have to show the police your ID if they ask for one. Use your best judgment, and consider looking up the laws for your state for more specific guidance.
Your phone (maybe): To protect your privacy and prevent surveillance, the best thing you can do is leave your phone at home. Consider using a secondary or burner phone instead. If you want to bring your phone, avoid using traditional phone calls and texts if at all possible. Signal is a secure, end-to-end encrypted messaging app that offers the option to delete messages after they’re sent. You should also disable biometric unlocking, like FaceID or fingerprint features, and use a six-digit passcode instead. If you do need to carry your primary phone, keep it turned off until you absolutely need to use it. This will make it harder for law enforcement to track your movements, since law enforcement officials can collect or purchase data that can tell them which devices were at a protest they’re monitoring, and who those devices belong to.
Cash: Just like your phone can leave digital breadcrumbs indicating your whereabouts, using your debit or credit card will make it easier for the authorities to track your movements. Instead, bring cash. Separate your bills; stash some in your bag, and keep some on your person, either in your shoe, your bra, your pants pockets, or somewhere else secure.
A power bank: If you or members of your group will have a phone, you need to make sure that you have a way to charge devices. Other protesters may need to charge their gear as well. If you don’t have one already, I really like this option from Anker. The company also makes another good power bank that’s a bit smaller.
Other things you may want: A cooling towel. Duct tape or gaffer’s tape. A flashlight or a headlamp. Ibuprofen. Goggles. Blister-prevention patches. Extra hair ties. A pen and paper. A Sharpie. A laser pointer. Bandages or other first aid supplies. Ear plugs. Saline solution. Extra face masks. A copy of emergency phone numbers and a card declaring necessary medical information that someone may need to know if you’re unable to tell them yourself (for example, if you have asthma or if you’re hard of hearing). Medications that need to be taken on a schedule (in a labeled prescription bottle if possible) with the understanding that you may be away from home much longer than anticipated.
Before You Leave
We asked some organizers and civil action experts about key things to understand about protesting. Here’s their advice.
Educate Yourself
Do not go to a protest without knowing what it is you’re fighting for. Don’t show up and ask someone there to educate you. If you’re an ally, do the work yourself and study as much as possible—not only about the actions you’re protesting but the context around them. You might know that reproductive rights are threatened, but do you know about the historical fight for access to abortion in America? There are several books you can read right now to gain more knowledge about this topic. If you’re pissed at Elon Musk, it helps to have a good grasp of all the ways the administration is dismantling federal agencies in this country.
Get in the Right Mindset
Mental preparation is important. Protesting can be physically grueling and emotional taxing. You may experience sheer joy. You might cry. You might get angry. You might get scared. Most likely, you’ll experience all of the above. Take the time to prepare yourself before heading out. Make sure to drink some water, apply sunscreen, and eat a meal.
Realize It’s Not About You
One of the primary reasons to attend a protest is not just to stick it to the man and to take a selfie while doing it, but to be present for others. Show up ready to listen. Be prepared to amplify what people are saying. If you aren’t comfortable with potentially physically intervening, shielding vulnerable protesters from police violence, and listening more than you speak, your efforts to be an ally are likely better spent elsewhere.
Tony Williams, a member of MPD150, a Minneapolis-based coalition that has studied the history of police activity and seeks police-free alternatives to community safety, shared crucial advice for anyone who is not personally impacted by current events, but who wants to attend a protest. (Of course, the dismantling of democracy in the US is something that affects all of us.)
“The most important thing to realize,” Williams says, “is that you’re showing up in solidarity with other people. It’s not your job to decide how things should go. It’s your job to show up and listen and be in support. Deprioritizing yourself is an incredibly important part of the experience.”
Know Your Rights
In the US, it’s entirely within your rights to peacefully demonstrate in public. The basic act of assembling and protesting the government’s actions is unquestionably protected, according to the First Amendment Coalition, a California nonprofit that’s committed to protecting freedom of speech. Also, as a general matter, “people have the right to film or otherwise document things that are happening in the public space,” says David Snyder, director of the FAC. “If police demand that you turn over your notes, I would say that you can assume they don’t have the right to seize that.”
That said, if it comes down to a matter of force and you are physically outmatched, you may have to weigh the risks to your immediate personal safety, potentially have your notes or phone stripped from you, and pursue legal action later on. Also, Snyder notes, the First Amendment to the Constitution does not protect protesters who engage in unlawful activity, which includes destroying property or assaulting other people.
